tri-ml / dd3d Goto Github PK

Official PyTorch implementation of DD3D: Is Pseudo-Lidar needed for Monocular 3D Object detection? (ICCV 2021), Dennis Park*, Rares Ambrus*, Vitor Guizilini, Jie Li, and Adrien Gaidon.

License: MIT License

Makefile 0.41% Dockerfile 0.80% Python 98.79%

computer-vision deep-learning pytorch

dd3d's People

Contributors

Stargazers

Watchers

Forkers

updating00 dennis-park-tri vitorguizilini-tri owen-liuyuxuan dstarer bigdatasciencegroup mostafa-mansour feifei-liu saulocatharino vivounicorn fungungun jlqzzz hiyyg liuwq0809 mfkiwl xaviolo99 revisitq demingwang wei-baldwin-zeng fyancy flipson dl19940602 shangbuhuan13 vivianyzw hatsunexym zhuokunyao sanmin0312 ksang collector-m kaixinbear lqs19881030 yanggui19891007 pyten percyjia95 sinead-li liyingyanucas betbetliu jinhuan-hit zggg1p jieli352 wuqiman fei4xu laokingshineuav kevinstoesser yijunwu weilaweila utc1205 zhumingxu yudai-kato-aisin bargul azuryl alfredqin omar-fouad aelahi23 sjg02122 towardsautonomy zhengfangwu wencheng256 edibhamzaarslan facadedevil andy-cheng xd1073321804 plum7ree leogift whuhxb monstre0731 jayvaghasiya xijunke deyh2020 seabird-go sprbull royliuyu r0bertr

dd3d's Issues

omninets train result is very low

[03/18 08:06:01 tridet.utils.hydra.callbacks]: Rank of current process: 0. World size: 8
[03/18 08:06:01 tridet.utils.setup]: Working Directory: /home/azuryl/dd3d_test/omni/dd3d/outputs/2023-03-18/08-05-55
[03/18 08:06:01 tridet.utils.setup]: Full config:
{
"WANDB": {
"ENABLED": false,
"DRYRUN": false,
"PROJECT": "dd3d",
"GROUP": null,
"TAGS": [
"kitti-val",
"dla34",
"bn"
]
},
"EVAL_ONLY": false,
"EVAL_ON_START": false,
"ONLY_REGISTER_DATASETS": false,
"OUTPUT_ROOT": "./outputs",
"SYNC_OUTPUT_DIR_S3": {
"ENABLED": false,
"ROOT_IN_S3": "???",
"PERIOD": 1000
},
"DATASET_ROOT": "/hdd1/datasets_left/",
"TMP_DIR": "/tmp/",
"DATASETS": {
"TRAIN": {
"NAME": "kitti_3d_train",
"CANONICAL_BOX3D_SIZES": [
[
1.61876949,
3.89154523,
1.52969237
],
[
0.62806586,
0.82038497,
1.76784787
],
[
0.56898187,
1.77149234,
1.7237099
],
[
1.9134491,
5.15499603,
2.18998422
],
[
2.61168401,
9.22692319,
3.36492722
],
[
0.5390196,
1.08098042,
1.28392158
],
[
2.36044838,
15.56991038,
3.5289238
],
[
1.24489164,
2.51495357,
1.61402478
]
],
"DATASET_MAPPER": "default",
"NUM_CLASSES": 5,
"MEAN_DEPTH_PER_LEVEL": [
32.594,
15.178,
8.424,
5.004,
4.662
],
"STD_DEPTH_PER_LEVEL": [
14.682,
7.139,
4.345,
2.399,
2.587
]
},
"TEST": {
"NAME": "kitti_3d_val",
"NUSC_SAMPLE_AGGREGATE_IN_INFERENCE": false,
"DATASET_MAPPER": "default"
}
},
"FE": {
"FPN": {
"IN_FEATURES": [
"level3",
"level4",
"level5"
],
"OUT_FEATURES": null,
"OUT_CHANNELS": 256,
"NORM": "FrozenBN",
"FUSE_TYPE": "sum"
},
"BUILDER": "build_fcos_dla_fpn_backbone_p67",
"BACKBONE": {
"NAME": "DLA-34",
"OUT_FEATURES": [
"level3",
"level4",
"level5"
],
"NORM": "FrozenBN"
},
"OUT_FEATURES": null
},
"DD3D": {
"IN_FEATURES": null,
"NUM_CLASSES": 5,
"FEATURE_LOCATIONS_OFFSET": "none",
"SIZES_OF_INTEREST": [
64,
128,
256,
512
],
"INFERENCE": {
"DO_NMS": true,
"DO_POSTPROCESS": true,
"DO_BEV_NMS": false,
"BEV_NMS_IOU_THRESH": 0.3,
"NUSC_SAMPLE_AGGREGATE": false
},
"FCOS2D": {
"_VERSION": "v2",
"NORM": "BN",
"NUM_CLS_CONVS": 4,
"NUM_BOX_CONVS": 4,
"USE_DEFORMABLE": false,
"USE_SCALE": true,
"BOX2D_SCALE_INIT_FACTOR": 1.0,
"LOSS": {
"ALPHA": 0.25,
"GAMMA": 2.0,
"LOC_LOSS_TYPE": "giou"
},
"INFERENCE": {
"THRESH_WITH_CTR": true,
"PRE_NMS_THRESH": 0.05,
"PRE_NMS_TOPK": 1000,
"POST_NMS_TOPK": 100,
"NMS_THRESH": 0.75
}
},
"FCOS3D": {
"NORM": "FrozenBN",
"NUM_CONVS": 4,
"USE_DEFORMABLE": false,
"USE_SCALE": true,
"DEPTH_SCALE_INIT_FACTOR": 0.3,
"PROJ_CTR_SCALE_INIT_FACTOR": 1.0,
"PER_LEVEL_PREDICTORS": false,
"SCALE_DEPTH_BY_FOCAL_LENGTHS": true,
"SCALE_DEPTH_BY_FOCAL_LENGTHS_FACTOR": 500.0,
"MEAN_DEPTH_PER_LEVEL": [
32.594,
15.178,
8.424,
5.004,
4.662
],
"STD_DEPTH_PER_LEVEL": [
14.682,
7.139,
4.345,
2.399,
2.587
],
"MIN_DEPTH": 0.1,
"MAX_DEPTH": 80.0,
"CANONICAL_BOX3D_SIZES": [
[
1.61876949,
3.89154523,
1.52969237
],
[
0.62806586,
0.82038497,
1.76784787
],
[
0.56898187,
1.77149234,
1.7237099
],
[
1.9134491,
5.15499603,
2.18998422
],
[
2.61168401,
9.22692319,
3.36492722
],
[
0.5390196,
1.08098042,
1.28392158
],
[
2.36044838,
15.56991038,
3.5289238
],
[
1.24489164,
2.51495357,
1.61402478
]
],
"CLASS_AGNOSTIC_BOX3D": false,
"PREDICT_ALLOCENTRIC_ROT": true,
"PREDICT_DISTANCE": false,
"LOSS": {
"SMOOTH_L1_BETA": 0.05,
"MAX_LOSS_PER_GROUP_DISENT": 20.0,
"CONF_3D_TEMPERATURE": 1.0,
"WEIGHT_BOX3D": 2.0,
"WEIGHT_CONF3D": 1.0
},
"PREPARE_TARGET": {
"CENTER_SAMPLE": true,
"POS_RADIUS": 1.5
}
}
},
"VIS": {
"DATALOADER_ENABLED": true,
"DATALOADER_PERIOD": 1000,
"DATALOADER_MAX_NUM_SAMPLES": 10,
"PREDICTIONS_ENABLED": true,
"PREDICTIONS_MAX_NUM_SAMPLES": 20,
"D2": {
"DATALOADER": {
"ENABLED": true,
"SCALE": 1.0,
"COLOR_MODE": "image"
},
"PREDICTIONS": {
"ENABLED": true,
"SCALE": 1.0,
"COLOR_MODE": "image",
"THRESHOLD": 0.4
}
},
"BOX3D": {
"DATALOADER": {
"ENABLED": true,
"SCALE": 1.0,
"RENDER_LABELS": true
},
"PREDICTIONS": {
"ENABLED": true,
"SCALE": 1.0,
"RENDER_LABELS": true,
"THRESHOLD": 0.5,
"MIN_DEPTH_CENTER": 0.0
}
}
},
"INPUT": {
"FORMAT": "BGR",
"AUG_ENABLED": true,
"RESIZE": {
"ENABLED": true,
"MIN_SIZE_TRAIN": [
288,
304,
320,
336,
352,
368,
384,
400,
416,
448,
480,
512,
544,
576
],
"MIN_SIZE_TRAIN_SAMPLING": "choice",
"MAX_SIZE_TRAIN": 10000,
"MIN_SIZE_TEST": 384,
"MAX_SIZE_TEST": 100000
},
"CROP": {
"ENABLED": false,
"TYPE": "relative_range",
"SIZE": [
0.9,
0.9
]
},
"RANDOM_FLIP": {
"ENABLED": true,
"HORIZONTAL": true,
"VERTICAL": false
},
"COLOR_JITTER": {
"ENABLED": true,
"BRIGHTNESS": [
0.2,
0.2
],
"SATURATION": [
0.2,
0.2
],
"CONTRAST": [
0.2,
0.2
]
}
},
"MODEL": {
"DEVICE": "cuda",
"META_ARCHITECTURE": "DD3D",
"PIXEL_MEAN": [
103.53,
116.28,
123.675
],
"PIXEL_STD": [
57.375,
57.12,
58.395
],
"CKPT": "",
"BOX2D_ON": true,
"BOX3D_ON": true,
"DEPTH_ON": false,
"backbone_with_fpn": null,
"width_mult": 1.0,
"depth_mult": 1.0
},
"DATALOADER": {
"TRAIN": {
"NUM_WORKERS": 12,
"FILTER_EMPTY_ANNOTATIONS": true,
"SAMPLER": "RepeatFactorTrainingSampler",
"REPEAT_THRESHOLD": 0.4,
"ASPECT_RATIO_GROUPING": false
},
"TEST": {
"NUM_WORKERS": 4,
"SAMPLER": "InferenceSampler"
}
},
"SOLVER": {
"IMS_PER_BATCH": 64,
"BASE_LR": 0.002,
"MOMENTUM": 0.9,
"NESTEROV": false,
"WEIGHT_DECAY": 0.0001,
"WEIGHT_DECAY_NORM": 0.0,
"BIAS_LR_FACTOR": 1.0,
"WEIGHT_DECAY_BIAS": 0.0001,
"GAMMA": 0.1,
"LR_SCHEDULER_NAME": "WarmupMultiStepLR",
"STEPS": [
21500,
24000
],
"WARMUP_FACTOR": 0.0001,
"WARMUP_ITERS": 2000,
"WARMUP_METHOD": "linear",
"CLIP_GRADIENTS": {
"ENABLED": false,
"CLIP_TYPE": "value",
"CLIP_VALUE": 1.0,
"NORM_TYPE": 2.0
},
"CHECKPOINT_PERIOD": 2000,
"MIXED_PRECISION_ENABLED": true,
"DDP_FIND_UNUSED_PARAMETERS": false,
"ACCUMULATE_GRAD_BATCHES": 1,
"SYNCBN_USE_LOCAL_WORKERS": false,
"MAX_ITER": 25000
},
"TEST": {
"ENABLED": true,
"EVAL_PERIOD": 2000,
"EVAL_ON_START": false,
"ADDITIONAL_EVAL_STEPS": [],
"IMS_PER_BATCH": 80,
"AUG": {
"ENABLED": true,
"MIN_SIZES": [
320,
384,
448,
512,
576
],
"MAX_SIZE": 100000,
"FLIP": true
}
},
"EVALUATORS": {
"KITTI3D": {
"IOU_THRESHOLDS": [
0.5,
0.7
],
"ONLY_PREPARE_SUBMISSION": false
}
},
"CKPT": "https://tri-ml-public.s3.amazonaws.com/github/dd3d/pretrained/depth_pretrained_omninet-small-3nxjur71.pth"
}
[03/18 08:06:01 tridet.data.datasets.kitti_3d]: KITTI-3D dataset(s): kitti_3d_train, kitti_3d_val
[03/18 08:06:08 tridet.data.build]: Creating D2 dataset: 15 batches on rank 0.
7%|▋ | 1/15 [00:02<00:28, 2.04s/it][2023-03-18 08:06:11,063][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:11,068][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:11,068][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:11,068][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:11,068][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:11,068][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:11,068][root][INFO] - Reducer buckets have been rebuilt in this iteration.
13%|█▎ | 2/15 [00:02<00:11, 1.10it/s][2023-03-18 08:06:11,115][root][INFO] - Reducer buckets have been rebuilt in this iteration.
100%|██████████| 15/15 [00:03<00:00, 4.94it/s][03/18 08:06:11 tridet.data.build]: Gathering D2 dataset dicts from all GPU workers...
100%|██████████| 15/15 [00:02<00:00, 5.65it/s][03/18 08:06:13 tridet.data.build]: Done (length=3712, took=1.7s).
WARNING [03/18 08:06:13 tridet.utils.coco]: Using previously cached COCO format annotations at '/tmp/kitti_3d_train_coco_format.json'. You need to clear the cache file if your dataset has been modified.
[03/18 08:06:13 tridet.data.datasets.kitti_3d.build]: COCO json file: /tmp/kitti_3d_train_coco_format.json

[03/18 08:06:20 tridet.data.build]: Creating D2 dataset: 15 batches on rank 0.
7%|▋ | 1/15 [00:01<00:27, 1.97s/it][2023-03-18 08:06:23,076][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:23,091][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:23,091][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:23,091][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:23,091][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:23,091][root][INFO] - Reducer buckets have been rebuilt in this iteration.
13%|█▎ | 2/15 [00:02<00:11, 1.13it/s][2023-03-18 08:06:23,136][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:23,136][root][INFO] - Reducer buckets have been rebuilt in this iteration.
100%|██████████| 15/15 [00:02<00:00, 5.69it/s][03/18 08:06:23 tridet.data.build]: Gathering D2 dataset dicts from all GPU workers...
100%|██████████| 15/15 [00:02<00:00, 5.67it/s][03/18 08:06:25 tridet.data.build]: Done (length=3769, took=1.5s).
WARNING [03/18 08:06:25 tridet.utils.coco]: Using previously cached COCO format annotations at '/tmp/kitti_3d_val_coco_format.json'. You need to clear the cache file if your dataset has been modified.
[03/18 08:06:25 tridet.data.datasets.kitti_3d.build]: COCO json file: /tmp/kitti_3d_val_coco_format.json
[03/18 08:06:25 tridet]: Registered 2 datasets:
kitti_3d_train
kitti_3d_val
[03/18 08:06:25 tridet.data.dataset_mappers.dataset_mapper]: Augmentations used in training: [ResizeShortestEdge(short_edge_length=[288, 304, 320, 336, 352, 368, 384, 400, 416, 448, 480, 512, 544, 576], max_size=10000, sample_style='choice'), RandomFlip(), RandomBrightness(intensity_min=0.8, intensity_max=1.2), RandomSaturation(intensity_min=0.8, intensity_max=1.2), RandomContrast(intensity_min=0.8, intensity_max=1.2)]
[03/18 08:06:25 d2.data.build]: Removed 0 images with no usable annotations. 3712 images left.
[03/18 08:06:26 d2.data.build]: Distribution of instances among all 5 categories:

category	#instances	category	#instances	category	#instances
Car	14357	Pedestrian	2207	Cyclist	734
Van	1297	Truck	488
total	19083
[03/18 08:06:26 d2.data.common]: Serializing 3712 elements to byte tensors and concatenating them all ...
[03/18 08:06:26 d2.data.common]: Serialized dataset takes 12.97 MiB
[03/18 08:06:26 tridet.data.build]: Using training sampler RepeatFactorTrainingSampler

2023-03-18 08:06:26,882][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 0
[2023-03-18 08:06:26,884][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 2
[2023-03-18 08:06:26,885][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 1
[2023-03-18 08:06:26,895][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 3
[2023-03-18 08:06:26,900][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 5
[2023-03-18 08:06:26,901][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 4
[2023-03-18 08:06:26,911][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 6
[2023-03-18 08:06:26,911][torch.distributed.distributed_c10d][INFO] - Added key: store_based_barrier_key:3 to store for rank: 7
[2023-03-18 08:06:26,911][torch.distributed.distributed_c10d][INFO] - Rank 6: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,911][torch.distributed.distributed_c10d][INFO] - Rank 5: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,912][torch.distributed.distributed_c10d][INFO] - Rank 7: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,912][torch.distributed.distributed_c10d][INFO] - Rank 4: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,914][torch.distributed.distributed_c10d][INFO] - Rank 0: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,916][torch.distributed.distributed_c10d][INFO] - Rank 3: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,916][torch.distributed.distributed_c10d][INFO] - Rank 2: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.
[2023-03-18 08:06:26,916][torch.distributed.distributed_c10d][INFO] - Rank 1: Completed store-based barrier for key:store_based_barrier_key:3 with 8 nodes.

[03/18 08:06:26 tridet]: Length of train dataset: 3712
[03/18 08:06:26 tridet]: Starting training

[2023-03-18 08:06:38,107][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,107][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,107][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,107][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,214][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,214][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,214][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[2023-03-18 08:06:38,297][root][INFO] - Reducer buckets have been rebuilt in this iteration.
[03/18 08:06:53 d2.utils.events]: iter: 20 total_loss: 313.7 loss_box2d_reg: 0.9839 loss_box3d_depth: 8.246 loss_box3d_proj_ctr: 0.5431 loss_box3d_quat: 2.739 loss_box3d_size: 1.337 loss_centerness: 0.829 loss_cls: 297.5 loss_conf3d: 1.106 lr: 1.9198e-05 max_mem: 5630M
[03/18 08:07:11 d2.utils.events]: eta: 6:05:01 iter: 40 total_loss: 16.35 loss_box2d_reg: 0.9829 loss_box3d_depth: 7.286 loss_box3d_proj_ctr: 0.4852 loss_box3d_quat: 2.643 loss_box3d_size: 1.003 loss_centerness: 0.7814 loss_cls: 2.374 loss_conf3d: 0.6532 lr: 3.9196e-05 max_mem: 5630M
[03/18 08:07:29 d2.utils.events]: eta: 6:15:22 iter: 60 total_loss: 15.62 loss_box2d_reg: 0.9841 loss_box3d_depth: 7.074 loss_box3d_proj_ctr: 0.4394 loss_box3d_quat: 2.56 loss_box3d_size: 0.8469 loss_centerness: 0.7516 loss_cls: 2.424 loss_conf3d: 0.4763 lr: 5.9194e-05 max_mem: 5630M
[03/18 08:07:45 d2.utils.events]: eta: 5:43:40 iter: 80 total_loss: 14.58 loss_box2d_reg: 0.9868 loss_box3d_depth: 6.905 loss_box3d_proj_ctr: 0.4186 loss_box3d_quat: 2.341 loss_box3d_size: 0.6894 loss_centerness: 0.7354 loss_cls: 2.056 loss_conf3d: 0.434 lr: 7.9192e-05 max_mem: 5630M
[03/18 08:08:02 d2.utils.events]: eta: 5:47:47 iter: 100 total_loss: 13.93 loss_box2d_reg: 0.9874 loss_box3d_depth: 6.831 loss_box3d_proj_ctr: 0.4029 loss_box3d_quat: 2.229 loss_box3d_size: 0.5956 loss_centerness: 0.7245 loss_cls: 1.662 loss_conf3d: 0.412 lr: 9.919e-05 max_mem: 5630M
[03/18 08:08:18 d2.utils.events]: eta: 5:30:20 iter: 120 total_loss: 13.23 loss_box2d_reg: 0.9885 loss_box3d_depth: 6.772 loss_box3d_proj_ctr: 0.4011 loss_box3d_quat: 2.174 loss_box3d_size: 0.5176 loss_centerness: 0.7142 loss_cls: 1.31 loss_conf3d: 0.4019 lr: 0.00011919 max_mem: 5630M
[03/18 08:08:35 d2.utils.events]: eta: 5:53:07 iter: 140 total_loss: 12.69 loss_box2d_reg: 0.9886 loss_box3d_depth: 6.417 loss_box3d_proj_ctr: 0.3833 loss_box3d_quat: 2.155 loss_box3d_size: 0.4419 loss_centerness: 0.7073 loss_cls: 1.047 loss_conf3d: 0.4043 lr: 0.00013919 max_mem: 5630M
[03/18 08:08:52 d2.utils.events]: eta: 5:50:52 iter: 160 total_loss: 12.11 loss_box2d_reg: 0.9847 loss_box3d_depth: 6.238 loss_box3d_proj_ctr: 0.3848 loss_box3d_quat: 2.105 loss_box3d_size: 0.4187 loss_centerness: 0.7041 loss_cls: 0.8634 loss_conf3d: 0.4038 lr: 0.00015918 max_mem: 5630M
[03/18 08:09:08 d2.utils.events]: eta: 5:36:01 iter: 180 total_loss: 11.85 loss_box2d_reg: 0.9781 loss_box3d_depth: 6.133 loss_box3d_proj_ctr: 0.3975 loss_box3d_quat: 2.138 loss_box3d_size: 0.4007 loss_centerness: 0.7013 loss_cls: 0.8141 loss_conf3d: 0.4097 lr: 0.00017918 max_mem: 5630M
[03/18 08:09:25 d2.utils.events]: eta: 5:49:53 iter: 200 total_loss: 11.32 loss_box2d_reg: 0.9691 loss_box3d_depth: 5.581 loss_box3d_proj_ctr: 0.3858 loss_box3d_quat: 2.057 loss_box3d_size: 0.3819 loss_centerness: 0.6958 loss_cls: 0.7868 loss_conf3d: 0.4348 lr: 0.00019918 max_mem: 5630M
[03/18 08:09:42 d2.utils.events]: eta: 5:50:24 iter: 220 total_loss: 10.9 loss_box2d_reg: 0.9443 loss_box3d_depth: 5.266 loss_box3d_proj_ctr: 0.3807 loss_box3d_quat: 2.119 loss_box3d_size: 0.358 loss_centerness: 0.6976 loss_cls: 0.7606 loss_conf3d: 0.436 lr: 0.00021918 max_mem: 5631M
[03/18 08:09:59 d2.utils.events]: eta: 5:36:50 iter: 240 total_loss: 10.65 loss_box2d_reg: 0.8994 loss_box3d_depth: 5.054 loss_box3d_proj_ctr: 0.3808 loss_box3d_quat: 2.041 loss_box3d_size: 0.3253 loss_centerness: 0.6955 loss_cls: 0.7061 loss_conf3d: 0.4434 lr: 0.00023918 max_mem: 5631M
[03/18 08:10:16 d2.utils.events]: eta: 6:03:12 iter: 260 total_loss: 10.58 loss_box2d_reg: 0.8071 loss_box3d_depth: 5.113 loss_box3d_proj_ctr: 0.3784 loss_box3d_quat: 2.126 loss_box3d_size: 0.3223 loss_centerness: 0.6966 loss_cls: 0.6927 loss_conf3d: 0.4359 lr: 0.00025917 max_mem: 5631M
[03/18 08:10:34 d2.utils.events]: eta: 6:12:29 iter: 280 total_loss: 9.967 loss_box2d_reg: 0.7079 loss_box3d_depth: 4.762 loss_box3d_proj_ctr: 0.3807 loss_box3d_quat: 2.071 loss_box3d_size: 0.3053 loss_centerness: 0.697 loss_cls: 0.6697 loss_conf3d: 0.4572 lr: 0.00027917 max_mem: 5631M
[03/18 08:10:52 d2.utils.events]: eta: 6:00:41 iter: 300 total_loss: 9.587 loss_box2d_reg: 0.6188 loss_box3d_depth: 4.493 loss_box3d_proj_ctr: 0.3809 loss_box3d_quat: 2.019 loss_box3d_size: 0.2915 loss_centerness: 0.6956 loss_cls: 0.6371 loss_conf3d: 0.4666 lr: 0.00029917 max_mem: 5631M
[03/18 08:11:09 d2.utils.events]: eta: 6:02:27 iter: 320 total_loss: 9.218 loss_box2d_reg: 0.5775 loss_box3d_depth: 4.211 loss_box3d_proj_ctr: 0.3782 loss_box3d_quat: 2.049 loss_box3d_size: 0.2695 loss_centerness: 0.6879 loss_cls: 0.6051 loss_conf3d: 0.4795 lr: 0.00031917 max_mem: 5631M
[03/18 08:11:26 d2.utils.events]: eta: 5:46:55 iter: 340 total_loss: 9.233 loss_box2d_reg: 0.5667 loss_box3d_depth: 4.277 loss_box3d_proj_ctr: 0.3803 loss_box3d_quat: 1.974 loss_box3d_size: 0.2613 loss_centerness: 0.6883 loss_cls: 0.5957 loss_conf3d: 0.4734 lr: 0.00033917 max_mem: 5631M
[03/18 08:11:42 d2.utils.events]: eta: 5:31:08 iter: 360 total_loss: 8.959 loss_box2d_reg: 0.5531 loss_box3d_depth: 3.969 loss_box3d_proj_ctr: 0.377 loss_box3d_quat: 1.966 loss_box3d_size: 0.2622 loss_centerness: 0.683 loss_cls: 0.5763 loss_conf3d: 0.4865 lr: 0.00035916 max_mem: 5631M
[03/18 08:11:59 d2.utils.events]: eta: 5:48:45 iter: 380 total_loss: 8.931 loss_box2d_reg: 0.5438 loss_box3d_depth: 3.986 loss_box3d_proj_ctr: 0.3742 loss_box3d_quat: 2.071 loss_box3d_size: 0.2466 loss_centerness: 0.6847 loss_cls: 0.5473 loss_conf3d: 0.4756 lr: 0.00037916 max_mem: 5631M
[03/18 08:12:16 d2.utils.events]: eta: 5:44:59 iter: 400 total_loss: 8.547 loss_box2d_reg: 0.53 loss_box3d_depth: 3.688 loss_box3d_proj_ctr: 0.3782 loss_box3d_quat: 2.032 loss_box3d_size: 0.2267 loss_centerness: 0.6795 loss_cls: 0.5292 loss_conf3d: 0.4964 lr: 0.00039916 max_mem: 5631M
[03/18 08:12:33 d2.utils.events]: eta: 5:41:55 iter: 420 total_loss: 8.449 loss_box2d_reg: 0.5203 loss_box3d_depth: 3.59 loss_box3d_proj_ctr: 0.3694 loss_box3d_quat: 2.06 loss_box3d_size: 0.2328 loss_centerness: 0.6779 loss_cls: 0.5298 loss_conf3d: 0.5062 lr: 0.00041916 max_mem: 5631M
[03/18 08:12:50 d2.utils.events]: eta: 5:58:25 iter: 440 total_loss: 8.701 loss_box2d_reg: 0.5144 loss_box3d_depth: 3.845 loss_box3d_proj_ctr: 0.3692 loss_box3d_quat: 2.052 loss_box3d_size: 0.2305 loss_centerness: 0.6765 loss_cls: 0.5357 loss_conf3d: 0.496 lr: 0.00043916 max_mem: 5631M
[03/18 08:13:08 d2.utils.events]: eta: 6:03:39 iter: 460 total_loss: 8.406 loss_box2d_reg: 0.5114 loss_box3d_depth: 3.646 loss_box3d_proj_ctr: 0.3705 loss_box3d_quat: 2.024 loss_box3d_size: 0.2223 loss_centerness: 0.6749 loss_cls: 0.5159 loss_conf3d: 0.5056 lr: 0.00045915 max_mem: 5631M
[03/18 08:13:25 d2.utils.events]: eta: 5:35:23 iter: 480 total_loss: 8.059 loss_box2d_reg: 0.495 loss_box3d_depth: 3.274 loss_box3d_proj_ctr: 0.3645 loss_box3d_quat: 1.989 loss_box3d_size: 0.2091 loss_centerness: 0.673 loss_cls: 0.4985 loss_conf3d: 0.5236 lr: 0.00047915 max_mem: 5631M
[03/18 08:13:43 d2.utils.events]: eta: 6:09:42 iter: 500 total_loss: 8.048 loss_box2d_reg: 0.4968 loss_box3d_depth: 3.301 loss_box3d_proj_ctr: 0.3631 loss_box3d_quat: 2.007 loss_box3d_size: 0.2059 loss_centerness: 0.6718 loss_cls: 0.4829 loss_conf3d: 0.5226 lr: 0.00049915 max_mem: 5631M
[03/18 08:14:02 d2.utils.events]: eta: 6:22:45 iter: 520 total_loss: 7.996 loss_box2d_reg: 0.4874 loss_box3d_depth: 3.196 loss_box3d_proj_ctr: 0.3659 loss_box3d_quat: 2.018 loss_box3d_size: 0.2089 loss_centerness: 0.6704 loss_cls: 0.482 loss_conf3d: 0.5192 lr: 0.00051915 max_mem: 5631M
[03/18 08:14:19 d2.utils.events]: eta: 5:52:08 iter: 540 total_loss: 8.394 loss_box2d_reg: 0.4856 loss_box3d_depth: 3.628 loss_box3d_proj_ctr: 0.3581 loss_box3d_quat: 2.013 loss_box3d_size: 0.2012 loss_centerness: 0.6691 loss_cls: 0.4796 loss_conf3d: 0.5054 lr: 0.00053915 max_mem: 5631M
[03/18 08:14:36 d2.utils.events]: eta: 5:56:55 iter: 560 total_loss: 7.878 loss_box2d_reg: 0.4753 loss_box3d_depth: 3.151 loss_box3d_proj_ctr: 0.3652 loss_box3d_quat: 2.065 loss_box3d_size: 0.1987 loss_centerness: 0.6684 loss_cls: 0.4594 loss_conf3d: 0.5235 lr: 0.00055914 max_mem: 5631M
[03/18 08:14:54 d2.utils.events]: eta: 5:59:40 iter: 580 total_loss: 7.752 loss_box2d_reg: 0.4706 loss_box3d_depth: 3.02 loss_box3d_proj_ctr: 0.3588 loss_box3d_quat: 2.003 loss_box3d_size: 0.1921 loss_centerness: 0.666 loss_cls: 0.4479 loss_conf3d: 0.5276 lr: 0.00057914 max_mem: 5631M
[03/18 08:15:11 d2.utils.events]: eta: 5:50:34 iter: 600 total_loss: 7.661 loss_box2d_reg: 0.4639 loss_box3d_depth: 2.968 loss_box3d_proj_ctr: 0.3508 loss_box3d_quat: 1.98 loss_box3d_size: 0.198 loss_centerness: 0.6653 loss_cls: 0.4515 loss_conf3d: 0.5431 lr: 0.00059914 max_mem: 5631M
[03/18 08:15:29 d2.utils.events]: eta: 6:02:12 iter: 620 total_loss: 7.837 loss_box2d_reg: 0.4569 loss_box3d_depth: 3.175 loss_box3d_proj_ctr: 0.3568 loss_box3d_quat: 1.978 loss_box3d_size: 0.1946 loss_centerness: 0.6636 loss_cls: 0.4433 loss_conf3d: 0.521 lr: 0.00061914 max_mem: 5631M
[03/18 08:15:47 d2.utils.events]: eta: 6:07:58 iter: 640 total_loss: 7.4 loss_box2d_reg: 0.4518 loss_box3d_depth: 2.776 loss_box3d_proj_ctr: 0.3484 loss_box3d_quat: 2.027 loss_box3d_size: 0.1891 loss_centerness: 0.6632 loss_cls: 0.4388 loss_conf3d: 0.5476 lr: 0.00063914 max_mem: 5631M
[03/18 08:16:05 d2.utils.events]: eta: 6:00:15 iter: 660 total_loss: 7.444 loss_box2d_reg: 0.4427 loss_box3d_depth: 2.746 loss_box3d_proj_ctr: 0.3471 loss_box3d_quat: 2.013 loss_box3d_size: 0.1857 loss_centerness: 0.6616 loss_cls: 0.4334 loss_conf3d: 0.5383 lr: 0.00065913 max_mem: 5631M
[03/18 08:16:23 d2.utils.events]: eta: 6:05:13 iter: 680 total_loss: 7.34 loss_box2d_reg: 0.4486 loss_box3d_depth: 2.729 loss_box3d_proj_ctr: 0.3551 loss_box3d_quat: 2.051 loss_box3d_size: 0.1861 loss_centerness: 0.6625 loss_cls: 0.4228 loss_conf3d: 0.5446 lr: 0.00067913 max_mem: 5631M
[03/18 08:16:41 d2.utils.events]: eta: 6:12:37 iter: 700 total_loss: 7.324 loss_box2d_reg: 0.4437 loss_box3d_depth: 2.778 loss_box3d_proj_ctr: 0.3369 loss_box3d_quat: 1.895 loss_box3d_size: 0.1822 loss_centerness: 0.6604 loss_cls: 0.421 loss_conf3d: 0.5549 lr: 0.00069913 max_mem: 5631M
[03/18 08:16:59 d2.utils.events]: eta: 5:53:25 iter: 720 total_loss: 7.247 loss_box2d_reg: 0.4349 loss_box3d_depth: 2.665 loss_box3d_proj_ctr: 0.3395 loss_box3d_quat: 2.071 loss_box3d_size: 0.1832 loss_centerness: 0.66 loss_cls: 0.4182 loss_conf3d: 0.5483 lr: 0.00071913 max_mem: 5631M
[03/18 08:17:18 d2.utils.events]: eta: 6:25:09 iter: 740 total_loss: 7.225 loss_box2d_reg: 0.427 loss_box3d_depth: 2.573 loss_box3d_proj_ctr: 0.3331 loss_box3d_quat: 2.058 loss_box3d_size: 0.1827 loss_centerness: 0.6577 loss_cls: 0.4156 loss_conf3d: 0.5472 lr: 0.00073913 max_mem: 5631M
[03/18 08:17:37 d2.utils.events]: eta: 6:18:08 iter: 760 total_loss: 7.479 loss_box2d_reg: 0.4307 loss_box3d_depth: 2.97 loss_box3d_proj_ctr: 0.3494 loss_box3d_quat: 2.002 loss_box3d_size: 0.1835 loss_centerness: 0.6579 loss_cls: 0.4124 loss_conf3d: 0.539 lr: 0.00075912 max_mem: 5631M
[03/18 08:17:55 d2.utils.events]: eta: 6:13:52 iter: 780 total_loss: 7.41 loss_box2d_reg: 0.4249 loss_box3d_depth: 2.805 loss_box3d_proj_ctr: 0.3338 loss_box3d_quat: 2.003 loss_box3d_size: 0.1762 loss_centerness: 0.6576 loss_cls: 0.4098 loss_conf3d: 0.5387 lr: 0.00077912 max_mem: 5631M
[03/18 08:18:14 d2.utils.events]: eta: 6:14:02 iter: 800 total_loss: 7.089 loss_box2d_reg: 0.4275 loss_box3d_depth: 2.618 loss_box3d_proj_ctr: 0.3316 loss_box3d_quat: 2.004 loss_box3d_size: 0.1814 loss_centerness: 0.6576 loss_cls: 0.4038 loss_conf3d: 0.5579 lr: 0.00079912 max_mem: 5632M
[03/18 08:18:32 d2.utils.events]: eta: 6:06:57 iter: 820 total_loss: 7.155 loss_box2d_reg: 0.4229 loss_box3d_depth: 2.65 loss_box3d_proj_ctr: 0.3232 loss_box3d_quat: 2.034 loss_box3d_size: 0.1791 loss_centerness: 0.6558 loss_cls: 0.4 loss_conf3d: 0.5496 lr: 0.00081912 max_mem: 5632M
[03/18 08:18:49 d2.utils.events]: eta: 5:45:03 iter: 840 total_loss: 6.939 loss_box2d_reg: 0.4146 loss_box3d_depth: 2.444 loss_box3d_proj_ctr: 0.3215 loss_box3d_quat: 2.012 loss_box3d_size: 0.1754 loss_centerness: 0.6562 loss_cls: 0.3967 loss_conf3d: 0.5615 lr: 0.00083912 max_mem: 5632M
[03/18 08:19:07 d2.utils.events]: eta: 6:01:24 iter: 860 total_loss: 7.042 loss_box2d_reg: 0.4196 loss_box3d_depth: 2.52 loss_box3d_proj_ctr: 0.3196 loss_box3d_quat: 1.966 loss_box3d_size: 0.1746 loss_centerness: 0.6566 loss_cls: 0.3916 loss_conf3d: 0.5585 lr: 0.00085911 max_mem: 5632M
[03/18 08:19:25 d2.utils.events]: eta: 6:08:57 iter: 880 total_loss: 7.178 loss_box2d_reg: 0.4112 loss_box3d_depth: 2.596 loss_box3d_proj_ctr: 0.3241 loss_box3d_quat: 1.964 loss_box3d_size: 0.1747 loss_centerness: 0.6557 loss_cls: 0.389 loss_conf3d: 0.5442 lr: 0.00087911 max_mem: 5632M
[03/18 08:19:43 d2.utils.events]: eta: 5:47:11 iter: 900 total_loss: 6.987 loss_box2d_reg: 0.4133 loss_box3d_depth: 2.505 loss_box3d_proj_ctr: 0.3177 loss_box3d_quat: 1.967 loss_box3d_size: 0.165 loss_centerness: 0.6551 loss_cls: 0.3895 loss_conf3d: 0.5635 lr: 0.00089911 max_mem: 5632M
[03/18 08:20:01 d2.utils.events]: eta: 6:12:12 iter: 920 total_loss: 6.741 loss_box2d_reg: 0.399 loss_box3d_depth: 2.223 loss_box3d_proj_ctr: 0.3157 loss_box3d_quat: 2.062 loss_box3d_size: 0.1692 loss_centerness: 0.6537 loss_cls: 0.3846 loss_conf3d: 0.572 lr: 0.00091911 max_mem: 5632M
[03/18 08:20:19 d2.utils.events]: eta: 6:05:16 iter: 940 total_loss: 6.783 loss_box2d_reg: 0.4011 loss_box3d_depth: 2.258 loss_box3d_proj_ctr: 0.315 loss_box3d_quat: 1.948 loss_box3d_size: 0.1725 loss_centerness: 0.6535 loss_cls: 0.3873 loss_conf3d: 0.5749 lr: 0.00093911 max_mem: 5632M
[03/18 08:20:37 d2.utils.events]: eta: 5:46:51 iter: 960 total_loss: 6.665 loss_box2d_reg: 0.3935 loss_box3d_depth: 2.254 loss_box3d_proj_ctr: 0.3167 loss_box3d_quat: 1.942 loss_box3d_size: 0.1703 loss_centerness: 0.6518 loss_cls: 0.3809 loss_conf3d: 0.5796 lr: 0.0009591 max_mem: 5632M
[03/18 08:20:56 d2.utils.events]: eta: 6:30:48 iter: 980 total_loss: 6.87 loss_box2d_reg: 0.3943 loss_box3d_depth: 2.313 loss_box3d_proj_ctr: 0.3179 loss_box3d_quat: 1.956 loss_box3d_size: 0.1642 loss_centerness: 0.6527 loss_cls: 0.38 loss_conf3d: 0.5673 lr: 0.0009791 max_mem: 5632M
[03/18 08:21:15 d2.utils.events]: eta: 6:18:15 iter: 1000 total_loss: 6.81 loss_box2d_reg: 0.3928 loss_box3d_depth: 2.418 loss_box3d_proj_ctr: 0.3128 loss_box3d_quat: 1.942 loss_box3d_size: 0.1678 loss_centerness: 0.6514 loss_cls: 0.3743 loss_conf3d: 0.5641 lr: 0.0009991 max_mem: 5632M
[03/18 08:21:38 d2.utils.events]: eta: 7:39:37 iter: 1020 total_loss: 6.905 loss_box2d_reg: 0.3801 loss_box3d_depth: 2.482 loss_box3d_proj_ctr: 0.3116 loss_box3d_quat: 1.942 loss_box3d_size: 0.1705 loss_centerness: 0.6507 loss_cls: 0.3672 loss_conf3d: 0.5583 lr: 0.0010191 max_mem: 5632M
[03/18 08:21:57 d2.utils.events]: eta: 6:12:08 iter: 1040 total_loss: 6.639 loss_box2d_reg: 0.3866 loss_box3d_depth: 2.269 loss_box3d_proj_ctr: 0.3061 loss_box3d_quat: 1.967 loss_box3d_size: 0.164 loss_centerness: 0.6509 loss_cls: 0.3658 loss_conf3d: 0.5779 lr: 0.0010391 max_mem: 5632M
[03/18 08:22:15 d2.utils.events]: eta: 6:11:48 iter: 1060 total_loss: 6.67 loss_box2d_reg: 0.3877 loss_box3d_depth: 2.281 loss_box3d_proj_ctr: 0.305 loss_box3d_quat: 1.972 loss_box3d_size: 0.1642 loss_centerness: 0.651 loss_cls: 0.3669 loss_conf3d: 0.5748 lr: 0.0010591 max_mem: 5632M
[03/18 08:22:33 d2.utils.events]: eta: 5:52:40 iter: 1080 total_loss: 6.735 loss_box2d_reg: 0.3809 loss_box3d_depth: 2.261 loss_box3d_proj_ctr: 0.3159 loss_box3d_quat: 1.955 loss_box3d_size: 0.1634 loss_centerness: 0.6518 loss_cls: 0.3699 loss_conf3d: 0.5751 lr: 0.0010791 max_mem: 5632M
[03/18 08:22:51 d2.utils.events]: eta: 5:57:27 iter: 1100 total_loss: 6.578 loss_box2d_reg: 0.3794 loss_box3d_depth: 2.198 loss_box3d_proj_ctr: 0.306 loss_box3d_quat: 1.953 loss_box3d_size: 0.1629 loss_centerness: 0.6504 loss_cls: 0.3662 loss_conf3d: 0.5751 lr: 0.0010991 max_mem: 5632M
[03/18 08:23:09 d2.utils.events]: eta: 5:59:44 iter: 1120 total_loss: 6.641 loss_box2d_reg: 0.3826 loss_box3d_depth: 2.281 loss_box3d_proj_ctr: 0.3073 loss_box3d_quat: 1.879 loss_box3d_size: 0.156 loss_centerness: 0.6519 loss_cls: 0.3631 loss_conf3d: 0.5691 lr: 0.0011191 max_mem: 5632M
[03/18 08:23:27 d2.utils.events]: eta: 6:01:38 iter: 1140 total_loss: 6.661 loss_box2d_reg: 0.3768 loss_box3d_depth: 2.177 loss_box3d_proj_ctr: 0.3096 loss_box3d_quat: 1.997 loss_box3d_size: 0.1643 loss_centerness: 0.6498 loss_cls: 0.3563 loss_conf3d: 0.5694 lr: 0.0011391 max_mem: 5632M
[03/18 08:23:46 d2.utils.events]: eta: 6:10:59 iter: 1160 total_loss: 6.691 loss_box2d_reg: 0.3709 loss_box3d_depth: 2.344 loss_box3d_proj_ctr: 0.3047 loss_box3d_quat: 1.934 loss_box3d_size: 0.1581 loss_centerness: 0.6487 loss_cls: 0.3619 loss_conf3d: 0.5679 lr: 0.0011591 max_mem: 5632M
[03/18 08:24:04 d2.utils.events]: eta: 6:04:55 iter: 1180 total_loss: 7.015 loss_box2d_reg: 0.3737 loss_box3d_depth: 2.619 loss_box3d_proj_ctr: 0.3101 loss_box3d_quat: 1.944 loss_box3d_size: 0.1604 loss_centerness: 0.6485 loss_cls: 0.3585 loss_conf3d: 0.5463 lr: 0.0011791 max_mem: 5632M
[03/18 08:24:23 d2.utils.events]: eta: 6:05:43 iter: 1200 total_loss: 6.973 loss_box2d_reg: 0.3769 loss_box3d_depth: 2.603 loss_box3d_proj_ctr: 0.3047 loss_box3d_quat: 1.943 loss_box3d_size: 0.1655 loss_centerness: 0.6503 loss_cls: 0.3687 loss_conf3d: 0.5576 lr: 0.0011991 max_mem: 5632M
[03/18 08:24:42 d2.utils.events]: eta: 6:18:14 iter: 1220 total_loss: 6.539 loss_box2d_reg: 0.375 loss_box3d_depth: 2.191 loss_box3d_proj_ctr: 0.3045 loss_box3d_quat: 1.939 loss_box3d_size: 0.1591 loss_centerness: 0.651 loss_cls: 0.3683 loss_conf3d: 0.5825 lr: 0.0012191 max_mem: 5632M
[03/18 08:25:02 d2.utils.events]: eta: 6:28:30 iter: 1240 total_loss: 6.376 loss_box2d_reg: 0.3713 loss_box3d_depth: 2.121 loss_box3d_proj_ctr: 0.2918 loss_box3d_quat: 1.792 loss_box3d_size: 0.1579 loss_centerness: 0.6496 loss_cls: 0.3558 loss_conf3d: 0.6053 lr: 0.0012391 max_mem: 5632M
[03/18 08:25:20 d2.utils.events]: eta: 6:05:09 iter: 1260 total_loss: 6.324 loss_box2d_reg: 0.3733 loss_box3d_depth: 2.279 loss_box3d_proj_ctr: 0.3035 loss_box3d_quat: 1.608 loss_box3d_size: 0.1526 loss_centerness: 0.6486 loss_cls: 0.3618 loss_conf3d: 0.6009 lr: 0.0012591 max_mem: 5632M
[03/18 08:25:39 d2.utils.events]: eta: 6:15:00 iter: 1280 total_loss: 6.361 loss_box2d_reg: 0.3683 loss_box3d_depth: 2.268 loss_box3d_proj_ctr: 0.31 loss_box3d_quat: 1.615 loss_box3d_size: 0.1639 loss_centerness: 0.6491 loss_cls: 0.3582 loss_conf3d: 0.6014 lr: 0.0012791 max_mem: 5632M
[03/18 08:25:58 d2.utils.events]: eta: 6:07:48 iter: 1300 total_loss: 6.099 loss_box2d_reg: 0.3677 loss_box3d_depth: 2.089 loss_box3d_proj_ctr: 0.3075 loss_box3d_quat: 1.525 loss_box3d_size: 0.165 loss_centerness: 0.6499 loss_cls: 0.3592 loss_conf3d: 0.6165 lr: 0.0012991 max_mem: 5632M
[03/18 08:26:16 d2.utils.events]: eta: 5:54:22 iter: 1320 total_loss: 6.157 loss_box2d_reg: 0.3632 loss_box3d_depth: 2.209 loss_box3d_proj_ctr: 0.3017 loss_box3d_quat: 1.495 loss_box3d_size: 0.1633 loss_centerness: 0.6475 loss_cls: 0.3518 loss_conf3d: 0.614 lr: 0.0013191 max_mem: 5632M
[03/18 08:26:35 d2.utils.events]: eta: 6:20:50 iter: 1340 total_loss: 6.085 loss_box2d_reg: 0.3638 loss_box3d_depth: 2.133 loss_box3d_proj_ctr: 0.3024 loss_box3d_quat: 1.456 loss_box3d_size: 0.1624 loss_centerness: 0.6482 loss_cls: 0.3557 loss_conf3d: 0.6216 lr: 0.0013391 max_mem: 5632M
[03/18 08:26:54 d2.utils.events]: eta: 6:10:26 iter: 1360 total_loss: 6.072 loss_box2d_reg: 0.368 loss_box3d_depth: 2.166 loss_box3d_proj_ctr: 0.3054 loss_box3d_quat: 1.49 loss_box3d_size: 0.1599 loss_centerness: 0.6474 loss_cls: 0.3537 loss_conf3d: 0.6147 lr: 0.0013591 max_mem: 5632M
[03/18 08:27:12 d2.utils.events]: eta: 6:05:03 iter: 1380 total_loss: 5.727 loss_box2d_reg: 0.3611 loss_box3d_depth: 1.988 loss_box3d_proj_ctr: 0.3012 loss_box3d_quat: 1.322 loss_box3d_size: 0.156 loss_centerness: 0.6484 loss_cls: 0.3463 loss_conf3d: 0.6342 lr: 0.0013791 max_mem: 5632M
[03/18 08:27:31 d2.utils.events]: eta: 6:09:04 iter: 1400 total_loss: 5.808 loss_box2d_reg: 0.3535 loss_box3d_depth: 2.093 loss_box3d_proj_ctr: 0.3004 loss_box3d_quat: 1.284 loss_box3d_size: 0.1502 loss_centerness: 0.6469 loss_cls: 0.3494 loss_conf3d: 0.6318 lr: 0.0013991 max_mem: 5632M
[03/18 08:27:50 d2.utils.events]: eta: 6:12:05 iter: 1420 total_loss: 5.784 loss_box2d_reg: 0.3496 loss_box3d_depth: 2.005 loss_box3d_proj_ctr: 0.2989 loss_box3d_quat: 1.343 loss_box3d_size: 0.1568 loss_centerness: 0.6455 loss_cls: 0.3433 loss_conf3d: 0.6302 lr: 0.0014191 max_mem: 5632M
[03/18 08:28:09 d2.utils.events]: eta: 6:04:13 iter: 1440 total_loss: 5.8 loss_box2d_reg: 0.3531 loss_box3d_depth: 2.089 loss_box3d_proj_ctr: 0.2929 loss_box3d_quat: 1.318 loss_box3d_size: 0.1598 loss_centerness: 0.6463 loss_cls: 0.3447 loss_conf3d: 0.6293 lr: 0.0014391 max_mem: 5632M
[03/18 08:28:28 d2.utils.events]: eta: 6:29:27 iter: 1460 total_loss: 5.891 loss_box2d_reg: 0.3608 loss_box3d_depth: 2.124 loss_box3d_proj_ctr: 0.292 loss_box3d_quat: 1.325 loss_box3d_size: 0.1578 loss_centerness: 0.6478 loss_cls: 0.3528 loss_conf3d: 0.6313 lr: 0.0014591 max_mem: 5632M
[03/18 08:28:48 d2.utils.events]: eta: 6:21:46 iter: 1480 total_loss: 5.804 loss_box2d_reg: 0.3537 loss_box3d_depth: 2.145 loss_box3d_proj_ctr: 0.3021 loss_box3d_quat: 1.24 loss_box3d_size: 0.1523 loss_centerness: 0.6464 loss_cls: 0.3407 loss_conf3d: 0.6331 lr: 0.0014791 max_mem: 5632M
[03/18 08:29:07 d2.utils.events]: eta: 6:08:08 iter: 1500 total_loss: 6.109 loss_box2d_reg: 0.3569 loss_box3d_depth: 2.405 loss_box3d_proj_ctr: 0.301 loss_box3d_quat: 1.284 loss_box3d_size: 0.1531 loss_centerness: 0.6459 loss_cls: 0.3489 loss_conf3d: 0.6149 lr: 0.0014991 max_mem: 5632M
[03/18 08:29:25 d2.utils.events]: eta: 6:05:07 iter: 1520 total_loss: 5.78 loss_box2d_reg: 0.3539 loss_box3d_depth: 2.094 loss_box3d_proj_ctr: 0.2956 loss_box3d_quat: 1.225 loss_box3d_size: 0.1588 loss_centerness: 0.6471 loss_cls: 0.3409 loss_conf3d: 0.6292 lr: 0.001519 max_mem: 5632M
[03/18 08:29:44 d2.utils.events]: eta: 5:57:01 iter: 1540 total_loss: 5.629 loss_box2d_reg: 0.3464 loss_box3d_depth: 2.034 loss_box3d_proj_ctr: 0.2877 loss_box3d_quat: 1.189 loss_box3d_size: 0.1499 loss_centerness: 0.6435 loss_cls: 0.3288 loss_conf3d: 0.6345 lr: 0.001539 max_mem: 5632M
[03/18 08:30:01 d2.utils.events]: eta: 5:48:09 iter: 1560 total_loss: 5.533 loss_box2d_reg: 0.3415 loss_box3d_depth: 1.951 loss_box3d_proj_ctr: 0.2918 loss_box3d_quat: 1.182 loss_box3d_size: 0.1534 loss_centerness: 0.6448 loss_cls: 0.3316 loss_conf3d: 0.637 lr: 0.001559 max_mem: 5632M
[03/18 08:30:21 d2.utils.events]: eta: 6:20:16 iter: 1580 total_loss: 6.106 loss_box2d_reg: 0.351 loss_box3d_depth: 2.564 loss_box3d_proj_ctr: 0.2969 loss_box3d_quat: 1.215 loss_box3d_size: 0.1525 loss_centerness: 0.6448 loss_cls: 0.3347 loss_conf3d: 0.6149 lr: 0.001579 max_mem: 5632M
[03/18 08:30:39 d2.utils.events]: eta: 6:00:58 iter: 1600 total_loss: 5.755 loss_box2d_reg: 0.3522 loss_box3d_depth: 2.135 loss_box3d_proj_ctr: 0.2831 loss_box3d_quat: 1.191 loss_box3d_size: 0.1524 loss_centerness: 0.646 loss_cls: 0.3369 loss_conf3d: 0.6341 lr: 0.001599 max_mem: 5632M
[03/18 08:30:57 d2.utils.events]: eta: 5:52:18 iter: 1620 total_loss: 5.558 loss_box2d_reg: 0.3479 loss_box3d_depth: 2.041 loss_box3d_proj_ctr: 0.2937 loss_box3d_quat: 1.187 loss_box3d_size: 0.1515 loss_centerness: 0.6454 loss_cls: 0.3347 loss_conf3d: 0.6389 lr: 0.001619 max_mem: 5632M
[03/18 08:31:16 d2.utils.events]: eta: 5:58:20 iter: 1640 total_loss: 5.535 loss_box2d_reg: 0.3432 loss_box3d_depth: 1.971 loss_box3d_proj_ctr: 0.2949 loss_box3d_quat: 1.18 loss_box3d_size: 0.1521 loss_centerness: 0.6447 loss_cls: 0.3311 loss_conf3d: 0.6382 lr: 0.001639 max_mem: 5632M
[03/18 08:31:35 d2.utils.events]: eta: 6:13:50 iter: 1660 total_loss: 5.641 loss_box2d_reg: 0.3409 loss_box3d_depth: 2.094 loss_box3d_proj_ctr: 0.2829 loss_box3d_quat: 1.119 loss_box3d_size: 0.1496 loss_centerness: 0.6433 loss_cls: 0.3285 loss_conf3d: 0.6339 lr: 0.001659 max_mem: 5632M
[03/18 08:31:53 d2.utils.events]: eta: 5:48:47 iter: 1680 total_loss: 5.577 loss_box2d_reg: 0.3472 loss_box3d_depth: 2.003 loss_box3d_proj_ctr: 0.2884 loss_box3d_quat: 1.109 loss_box3d_size: 0.1493 loss_centerness: 0.6456 loss_cls: 0.3313 loss_conf3d: 0.6348 lr: 0.001679 max_mem: 5632M
[03/18 08:32:13 d2.utils.events]: eta: 6:26:15 iter: 1700 total_loss: 5.497 loss_box2d_reg: 0.3394 loss_box3d_depth: 1.937 loss_box3d_proj_ctr: 0.2919 loss_box3d_quat: 1.165 loss_box3d_size: 0.1506 loss_centerness: 0.6444 loss_cls: 0.3255 loss_conf3d: 0.6378 lr: 0.001699 max_mem: 5632M
[03/18 08:32:33 d2.utils.events]: eta: 6:22:42 iter: 1720 total_loss: 5.457 loss_box2d_reg: 0.3388 loss_box3d_depth: 1.955 loss_box3d_proj_ctr: 0.2799 loss_box3d_quat: 1.086 loss_box3d_size: 0.1483 loss_centerness: 0.644 loss_cls: 0.3155 loss_conf3d: 0.6394 lr: 0.001719 max_mem: 5632M
[03/18 08:32:51 d2.utils.events]: eta: 5:59:36 iter: 1740 total_loss: 5.458 loss_box2d_reg: 0.3355 loss_box3d_depth: 1.962 loss_box3d_proj_ctr: 0.285 loss_box3d_quat: 1.095 loss_box3d_size: 0.1466 loss_centerness: 0.6434 loss_cls: 0.3263 loss_conf3d: 0.6384 lr: 0.001739 max_mem: 5632M
[03/18 08:33:10 d2.utils.events]: eta: 6:11:12 iter: 1760 total_loss: 5.344 loss_box2d_reg: 0.3334 loss_box3d_depth: 1.909 loss_box3d_proj_ctr: 0.2814 loss_box3d_quat: 1.049 loss_box3d_size: 0.1418 loss_centerness: 0.6441 loss_cls: 0.3185 loss_conf3d: 0.6423 lr: 0.001759 max_mem: 5632M
[03/18 08:33:29 d2.utils.events]: eta: 6:07:56 iter: 1780 total_loss: 5.221 loss_box2d_reg: 0.3335 loss_box3d_depth: 1.867 loss_box3d_proj_ctr: 0.2846 loss_box3d_quat: 1.031 loss_box3d_size: 0.1468 loss_centerness: 0.6424 loss_cls: 0.3144 loss_conf3d: 0.6463 lr: 0.001779 max_mem: 5632M
[03/18 08:33:47 d2.utils.events]: eta: 5:46:48 iter: 1800 total_loss: 5.386 loss_box2d_reg: 0.3327 loss_box3d_depth: 1.915 loss_box3d_proj_ctr: 0.2877 loss_box3d_quat: 1.108 loss_box3d_size: 0.1585 loss_centerness: 0.6427 loss_cls: 0.3144 loss_conf3d: 0.6437 lr: 0.001799 max_mem: 5632M
[03/18 08:34:06 d2.utils.events]: eta: 5:50:54 iter: 1820 total_loss: 5.412 loss_box2d_reg: 0.3317 loss_box3d_depth: 2.006 loss_box3d_proj_ctr: 0.2895 loss_box3d_quat: 1.022 loss_box3d_size: 0.1489 loss_centerness: 0.6423 loss_cls: 0.3128 loss_conf3d: 0.6448 lr: 0.001819 max_mem: 5632M
[03/18 08:34:24 d2.utils.events]: eta: 6:02:58 iter: 1840 total_loss: 5.868 loss_box2d_reg: 0.3389 loss_box3d_depth: 2.383 loss_box3d_proj_ctr: 0.2912 loss_box3d_quat: 1.055 loss_box3d_size: 0.1456 loss_centerness: 0.6443 loss_cls: 0.3225 loss_conf3d: 0.6344 lr: 0.001839 max_mem: 5632M
[03/18 08:34:42 d2.utils.events]: eta: 5:39:11 iter: 1860 total_loss: 5.477 loss_box2d_reg: 0.3404 loss_box3d_depth: 2.04 loss_box3d_proj_ctr: 0.2862 loss_box3d_quat: 1.025 loss_box3d_size: 0.148 loss_centerness: 0.6442 loss_cls: 0.3221 loss_conf3d: 0.6452 lr: 0.001859 max_mem: 5632M
[03/18 08:35:00 d2.utils.events]: eta: 5:49:06 iter: 1880 total_loss: 5.558 loss_box2d_reg: 0.3291 loss_box3d_depth: 2.17 loss_box3d_proj_ctr: 0.2895 loss_box3d_quat: 0.989 loss_box3d_size: 0.1437 loss_centerness: 0.6424 loss_cls: 0.3169 loss_conf3d: 0.638 lr: 0.001879 max_mem: 5632M
[03/18 08:35:19 d2.utils.events]: eta: 6:08:45 iter: 1900 total_loss: 5.253 loss_box2d_reg: 0.3289 loss_box3d_depth: 1.898 loss_box3d_proj_ctr: 0.2903 loss_box3d_quat: 0.9442 loss_box3d_size: 0.1506 loss_centerness: 0.6417 loss_cls: 0.3157 loss_conf3d: 0.6479 lr: 0.001899 max_mem: 5632M
[03/18 08:35:37 d2.utils.events]: eta: 5:41:41 iter: 1920 total_loss: 5.343 loss_box2d_reg: 0.3279 loss_box3d_depth: 1.974 loss_box3d_proj_ctr: 0.2896 loss_box3d_quat: 0.9662 loss_box3d_size: 0.1469 loss_centerness: 0.6409 loss_cls: 0.313 loss_conf3d: 0.6459 lr: 0.001919 max_mem: 5632M
[03/18 08:35:56 d2.utils.events]: eta: 6:06:53 iter: 1940 total_loss: 5.245 loss_box2d_reg: 0.3287 loss_box3d_depth: 1.981 loss_box3d_proj_ctr: 0.2698 loss_box3d_quat: 0.9367 loss_box3d_size: 0.1429 loss_centerness: 0.6427 loss_cls: 0.3073 loss_conf3d: 0.6487 lr: 0.001939 max_mem: 5632M
[03/18 08:36:16 d2.utils.events]: eta: 6:22:55 iter: 1960 total_loss: 5.244 loss_box2d_reg: 0.3229 loss_box3d_depth: 1.9 loss_box3d_proj_ctr: 0.2822 loss_box3d_quat: 1.009 loss_box3d_size: 0.1435 loss_centerness: 0.6422 loss_cls: 0.3102 loss_conf3d: 0.6472 lr: 0.001959 max_mem: 5632M
[03/18 08:36:34 d2.utils.events]: eta: 5:43:01 iter: 1980 total_loss: 5.142 loss_box2d_reg: 0.3235 loss_box3d_depth: 1.81 loss_box3d_proj_ctr: 0.2837 loss_box3d_quat: 0.945 loss_box3d_size: 0.1464 loss_centerness: 0.6411 loss_cls: 0.3118 loss_conf3d: 0.6522 lr: 0.001979 max_mem: 5632M
[03/18 08:36:52 d2.utils.events]: eta: 5:50:43 iter: 2000 total_loss: 5.204 loss_box2d_reg: 0.3229 loss_box3d_depth: 1.922 loss_box3d_proj_ctr: 0.2814 loss_box3d_quat: 0.9451 loss_box3d_size: 0.1452 loss_centerness: 0.6421 loss_cls: 0.3056 loss_conf3d: 0.6506 lr: 0.001999 max_mem: 5632M
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
[03/18 08:37:00 fvcore.common.checkpoint]: Saving checkpoint to ./model_0001999.pth
[03/18 08:37:01 tridet.data.dataset_mappers.dataset_mapper]: Augmentations used in training: [ResizeShortestEdge(short_edge_length=(384, 384), max_size=100000, sample_style='choice')]
[03/18 08:37:01 d2.data.common]: Serializing 3769 elements to byte tensors and concatenating them all ...
[03/18 08:37:01 d2.data.common]: Serialized dataset takes 13.38 MiB
[03/18 08:37:01 tridet.data.build]: Using test sampler InferenceSampler
[03/18 08:37:01 d2.evaluation.evaluator]: Start inference on 48 batches
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/aten/src/ATen/native/TensorShape.cpp:2157.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
/home/azuryl/anaconda3/envs/dd3d/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:134: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:87: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
[03/18 08:37:21 d2.evaluation.evaluator]: Inference done 11/48. Dataloading: 0.0009 s/iter. Inference: 0.3040 s/iter. Eval: 1.2617 s/iter. Total: 1.5665 s/iter. ETA=0:00:57
[03/18 08:37:27 d2.evaluation.evaluator]: Inference done 15/48. Dataloading: 0.0009 s/iter. Inference: 0.3116 s/iter. Eval: 1.2645 s/iter. Total: 1.5771 s/iter. ETA=0:00:52
[03/18 08:37:32 d2.evaluation.evaluator]: Inference done 18/48. Dataloading: 0.0010 s/iter. Inference: 0.3153 s/iter. Eval: 1.3108 s/iter. Total: 1.6273 s/iter. ETA=0:00:48
[03/18 08:37:39 d2.evaluation.evaluator]: Inference done 22/48. Dataloading: 0.0010 s/iter. Inference: 0.3119 s/iter. Eval: 1.3120 s/iter. Total: 1.6249 s/iter. ETA=0:00:42
[03/18 08:37:45 d2.evaluation.evaluator]: Inference done 26/48. Dataloading: 0.0010 s/iter. Inference: 0.3091 s/iter. Eval: 1.3096 s/iter. Total: 1.6198 s/iter. ETA=0:00:35
[03/18 08:37:52 d2.evaluation.evaluator]: Inference done 30/48. Dataloading: 0.0010 s/iter. Inference: 0.3094 s/iter. Eval: 1.3118 s/iter. Total: 1.6223 s/iter. ETA=0:00:29
[03/18 08:37:58 d2.evaluation.evaluator]: Inference done 34/48. Dataloading: 0.0010 s/iter. Inference: 0.3091 s/iter. Eval: 1.3145 s/iter. Total: 1.6248 s/iter. ETA=0:00:22
[03/18 08:38:04 d2.evaluation.evaluator]: Inference done 38/48. Dataloading: 0.0010 s/iter. Inference: 0.3088 s/iter. Eval: 1.3051 s/iter. Total: 1.6150 s/iter. ETA=0:00:16
[03/18 08:38:10 d2.evaluation.evaluator]: Inference done 42/48. Dataloading: 0.0010 s/iter. Inference: 0.3082 s/iter. Eval: 1.2912 s/iter. Total: 1.6005 s/iter. ETA=0:00:09
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
/home/azuryl/dd3d_test/omni/dd3d/scripts/tridet/structures/boxes3d.py:210: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1640811805959/work/torch/csrc/utils/tensor_new.cpp:201.)
quats = torch.as_tensor(quats, dtype=torch.float32, device=device)
[03/18 08:38:16 d2.evaluation.evaluator]: Inference done 46/48. Dataloading: 0.0010 s/iter. Inference: 0.3067 s/iter. Eval: 1.2753 s/iter. Total: 1.5832 s/iter. ETA=0:00:03
[03/18 08:38:18 d2.evaluation.evaluator]: Total inference time: 0:01:06.936973 (1.556674 s / iter per device, on 8 devices)
[03/18 08:38:18 d2.evaluation.evaluator]: Total inference pure compute time: 0:00:13 (0.304431 s / iter per device, on 8 devices)
################coco_evaluation.py evaluate
[03/18 08:38:22 d2.evaluation.coco_evaluation]: @@@@@@@@@@@@@@@@@###############Preparing results for COCO format ...
[03/18 08:38:22 d2.evaluation.coco_evaluation]: Saving results to /home/azuryl/dd3d_test/omni/dd3d/outputs/2023-03-18/08-05-55/inference/step0002000/kitti_3d_val/coco_instances_results.json
[03/18 08:38:24 d2.evaluation.coco_evaluation]: Evaluating predictions with unofficial COCO API...
Loading and preparing results...
DONE (t=1.02s)
creating index...
index created!
[03/18 08:38:25 d2.evaluation.fast_eval_api]: Evaluate annotation type bbox
[03/18 08:38:29 d2.evaluation.fast_eval_api]: COCOeval_opt.evaluate() finished in 3.81 seconds.
[03/18 08:38:29 d2.evaluation.fast_eval_api]: Accumulating evaluation results...
[03/18 08:38:30 d2.evaluation.fast_eval_api]: COCOeval_opt.accumulate() finished in 1.12 seconds.
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.062
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.148
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.043
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.067
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.065
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.081
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.083
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.206
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.254
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.135
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.273
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.300
[03/18 08:38:30 d2.evaluation.coco_evaluation]: Evaluation results for bbox:

AP	AP50	AP75	APs	APm	APl
6.195	14.801	4.259	6.657	6.480	8.127
[03/18 08:38:30 d2.evaluation.coco_evaluation]: Per-category bbox AP:
category	AP	category	AP	category	AP
:-----------	:-------	:-----------	:------	:-----------	:------
Car	22.022	Pedestrian	3.769	Cyclist	0.333
Van	2.425	Truck	2.424
[03/18 08:39:36 tridet]: Running prediction visualizer: d2_visualizer
[03/18 08:39:38 tridet.visualizers.d2_visualizer]: Found 2D detection predictions (bbox2d and/or mask2d) for 3769 images.
100%	██████████	20/20 [00:04<00:00, 4.70it/s][03/18 08:39:43 tridet]: Running prediction visualizer: box3d_visualizer

100%|██████████| 20/20 [00:07<00:00, 2.74it/s][03/18 08:40:01 tridet]: Evaluation results for kitti_3d_val in csv format:
[03/18 08:40:01 tridet.utils.train]: Test results:

metric	value
kitti_3d_val/bbox/AP	6.19459
kitti_3d_val/bbox/AP50	14.8013
kitti_3d_val/bbox/AP75	4.25872
kitti_3d_val/bbox/APs	6.657
kitti_3d_val/bbox/APm	6.47991
kitti_3d_val/bbox/APl	8.12651
kitti_3d_val/bbox/AP-Car	22.0222
kitti_3d_val/bbox/AP-Pedestrian	3.76924
kitti_3d_val/bbox/AP-Cyclist	0.332992
kitti_3d_val/bbox/AP-Van	2.42476
kitti_3d_val/bbox/AP-Truck	2.4238
kitti_3d_val/kitti_box3d_r40/Car_Easy_0.5	3.43969
kitti_3d_val/kitti_box3d_r40/Car_Easy_0.7	0.0141254
kitti_3d_val/kitti_box3d_r40/Car_Moderate_0.5	2.88214
kitti_3d_val/kitti_box3d_r40/Car_Moderate_0.7	0.00878241
kitti_3d_val/kitti_box3d_r40/Car_Hard_0.5	2.60383
kitti_3d_val/kitti_box3d_r40/Car_Hard_0.7	0.00702752
kitti_3d_val/kitti_box3d_r40/Pedestrian_Easy_0.5	0.0676474
kitti_3d_val/kitti_box3d_r40/Pedestrian_Easy_0.7	0
kitti_3d_val/kitti_box3d_r40/Pedestrian_Moderate_0.5	0.0401736
kitti_3d_val/kitti_box3d_r40/Pedestrian_Moderate_0.7	0
kitti_3d_val/kitti_box3d_r40/Pedestrian_Hard_0.5	0.0235035
kitti_3d_val/kitti_box3d_r40/Pedestrian_Hard_0.7	0
kitti_3d_val/kitti_box3d_r40/Cyclist_Easy_0.5	0.001665
kitti_3d_val/kitti_box3d_r40/Cyclist_Easy_0.7	0
kitti_3d_val/kitti_box3d_r40/Cyclist_Moderate_0.5	0
kitti_3d_val/kitti_box3d_r40/Cyclist_Moderate_0.7	0
kitti_3d_val/kitti_box3d_r40/Cyclist_Hard_0.5	0
kitti_3d_val/kitti_box3d_r40/Cyclist_Hard_0.7	0
kitti_3d_val/kitti_box3d_r40/Van_Easy_0.5	0.0625547
kitti_3d_val/kitti_box3d_r40/Van_Easy_0.7	0
kitti_3d_val/kitti_box3d_r40/Van_Moderate_0.5	0.012554
kitti_3d_val/kitti_box3d_r40/Van_Moderate_0.7	0
kitti_3d_val/kitti_box3d_r40/Van_Hard_0.5	0.015326
kitti_3d_val/kitti_box3d_r40/Van_Hard_0.7	0
kitti_3d_val/kitti_box3d_r40/Truck_Easy_0.5	0.0015949
kitti_3d_val/kitti_box3d_r40/Truck_Easy_0.7	0
kitti_3d_val/kitti_box3d_r40/Truck_Moderate_0.5	0
kitti_3d_val/kitti_box3d_r40/Truck_Moderate_0.7	0
kitti_3d_val/kitti_box3d_r40/Truck_Hard_0.5	0
kitti_3d_val/kitti_box3d_r40/Truck_Hard_0.7	0
kitti_3d_val/kitti_bev_r40/Car_Easy_0.5	6.40887
kitti_3d_val/kitti_bev_r40/Car_Easy_0.7	0.147282
kitti_3d_val/kitti_bev_r40/Car_Moderate_0.5	5.14902
kitti_3d_val/kitti_bev_r40/Car_Moderate_0.7	0.167439
kitti_3d_val/kitti_bev_r40/Car_Hard_0.5	4.61539
kitti_3d_val/kitti_bev_r40/Car_Hard_0.7	0.143622
kitti_3d_val/kitti_bev_r40/Pedestrian_Easy_0.5	0.235183
kitti_3d_val/kitti_bev_r40/Pedestrian_Easy_0.7	0
kitti_3d_val/kitti_bev_r40/Pedestrian_Moderate_0.5	0.16232
kitti_3d_val/kitti_bev_r40/Pedestrian_Moderate_0.7	0
kitti_3d_val/kitti_bev_r40/Pedestrian_Hard_0.5	0.145262
kitti_3d_val/kitti_bev_r40/Pedestrian_Hard_0.7	0
kitti_3d_val/kitti_bev_r40/Cyclist_Easy_0.5	0.00284437
kitti_3d_val/kitti_bev_r40/Cyclist_Easy_0.7	0
kitti_3d_val/kitti_bev_r40/Cyclist_Moderate_0.5	0.00326675
kitti_3d_val/kitti_bev_r40/Cyclist_Moderate_0.7	0
kitti_3d_val/kitti_bev_r40/Cyclist_Hard_0.5	0.00281233
kitti_3d_val/kitti_bev_r40/Cyclist_Hard_0.7	0
kitti_3d_val/kitti_bev_r40/Van_Easy_0.5	0.208593
kitti_3d_val/kitti_bev_r40/Van_Easy_0.7	0
kitti_3d_val/kitti_bev_r40/Van_Moderate_0.5	0.0855819
kitti_3d_val/kitti_bev_r40/Van_Moderate_0.7	0
kitti_3d_val/kitti_bev_r40/Van_Hard_0.5	0.0675817
kitti_3d_val/kitti_bev_r40/Van_Hard_0.7	0
kitti_3d_val/kitti_bev_r40/Truck_Easy_0.5	0.0705497
kitti_3d_val/kitti_bev_r40/Truck_Easy_0.7	0
kitti_3d_val/kitti_bev_r40/Truck_Moderate_0.5	0.0119722
kitti_3d_val/kitti_bev_r40/Truck_Moderate_0.7	0
kitti_3d_val/kitti_bev_r40/Truck_Hard_0.5	0.012178
kitti_3d_val/kitti_bev_r40/Truck_Hard_0.7	0
[03/18 08:40:20 d2.utils.events]: eta: 2 days, 18:21:58 iter: 2020 total_loss: 5.169 loss_box2d_reg: 0.3253 loss_box3d_depth: 1.877 loss_box3d_proj_ctr: 0.2758 loss_box3d_quat: 0.9252 loss_box3d_size: 0.1428 loss_centerness: 0.6408 loss_cls: 0.309 loss_conf3d: 0.6467 lr: 0.002 max_mem: 5632M
[03/18 08:40:37 d2.utils.events]: eta: 5:32:19 iter: 2040 total_loss: 5.002 loss_box2d_reg: 0.3238 loss_box3d_depth: 1.718 loss_box3d_proj_ctr: 0.2791 loss_box3d_quat: 0.8881 loss_box3d_size: 0.1407 loss_centerness: 0.642 loss_cls: 0.3085 loss_conf3d: 0.6524 lr: 0.002 max_mem: 5632M
[03/18 08:40:56 d2.utils.events]: eta: 5:44:39 iter: 2060 total_loss: 5.184 loss_box2d_reg: 0.3183 loss_box3d_depth: 1.875 loss_box3d_proj_ctr: 0.2844 loss_box3d_quat: 0.9504 loss_box3d_size: 0.1463 loss_centerness: 0.641 loss_cls: 0.3013 loss_conf3d: 0.6473 lr: 0.002 max_mem: 5632M
[03/18 08:41:15 d2.utils.events]: eta: 6:03:07 iter: 2080 total_loss: 5.115 loss_box2d_reg: 0.3204 loss_box3d_depth: 1.913 loss_box3d_proj_ctr: 0.2668 loss_box3d_quat: 0.896 loss_box3d_size: 0.137 loss_centerness: 0.6414 loss_cls: 0.3008 loss_conf3d: 0.6513 lr: 0.002 max_mem: 5632M
[03/18 08:41:33 d2.utils.events]: eta: 5:46:34 iter: 2100 total_loss: 4.99 loss_box2d_reg: 0.3176 loss_box3d_depth: 1.789 loss_box3d_proj_ctr: 0.2682 loss_box3d_quat: 0.8719 loss_box3d_size: 0.1449 loss_centerness: 0.6403 loss_cls: 0.3038 loss_conf3d: 0.6554 lr: 0.002 max_mem: 5632M
[03/18 08:41:52 d2.utils.events]: eta: 6:03:21 iter: 2120 total_loss: 4.894 loss_box2d_reg: 0.3164 loss_box3d_depth: 1.667 loss_box3d_proj_ctr: 0.2807 loss_box3d_quat: 0.9206 loss_box3d_size: 0.1399 loss_centerness: 0.6404 loss_cls: 0.294 loss_conf3d: 0.6571 lr: 0.002 max_mem: 5632M
[03/18 08:42:11 d2.utils.events]: eta: 5:58:01 iter: 2140 total_loss: 4.832 loss_box2d_reg: 0.3119 loss_box3d_depth: 1.683 loss_box3d_proj_ctr: 0.2716 loss_box3d_quat: 0.8633 loss_box3d_size: 0.1407 loss_centerness: 0.6405 loss_cls: 0.2968 loss_conf3d: 0.6586 lr: 0.002 max_mem: 5632M
[03/18 08:42:28 d2.utils.events]: eta: 5:33:51 iter: 2160 total_loss: 4.857 loss_box2d_reg: 0.3136 loss_box3d_depth: 1.718 loss_box3d_proj_ctr: 0.2676 loss_box3d_quat: 0.8268 loss_box3d_size: 0.1383 loss_centerness: 0.6403 loss_cls: 0.2959 loss_conf3d: 0.655 lr: 0.002 max_mem: 5632M
[03/18 08:42:47 d2.utils.events]: eta: 6:01:27 iter: 2180 total_loss: 4.784 loss_box2d_reg: 0.3121 loss_box3d_depth: 1.679 loss_box3d_proj_ctr: 0.2706 loss_box3d_quat: 0.8455 loss_box3d_size: 0.1403 loss_centerness: 0.6393 loss_cls: 0.2948 loss_conf3d: 0.6529 lr: 0.002 max_mem: 5632M
[03/18 08:43:06 d2.utils.events]: eta: 5:49:53 iter: 2200 total_loss: 4.681 loss_box2d_reg: 0.3097 loss_box3d_depth: 1.515 loss_box3d_proj_ctr: 0.2738 loss_box3d_quat: 0.8181 loss_box3d_size: 0.1404 loss_centerness: 0.6396 loss_cls: 0.2876 loss_conf3d: 0.6565 lr: 0.002 max_mem: 5632M
[03/18 08:43:23 d2.utils.events]: eta: 5:32:49 iter: 2220 total_loss: 4.815 loss_box2d_reg: 0.3053 loss_box3d_depth: 1.639 loss_box3d_proj_ctr: 0.2698 loss_box3d_quat: 0.8319 loss_box3d_size: 0.1362 loss_centerness: 0.6396 loss_cls: 0.29 loss_conf3d: 0.6553 lr: 0.002 max_mem: 5632M
[03/18 08:43:42 d2.utils.events]: eta: 5:57:07 iter: 2240 total_loss: 4.928 loss_box2d_reg: 0.3047 loss_box3d_depth: 1.749 loss_box3d_proj_ctr: 0.2722 loss_box3d_quat: 0.8189 loss_box3d_size: 0.1449 loss_centerness: 0.6394 loss_cls: 0.2876 loss_conf3d: 0.6539 lr: 0.002 max_mem: 5632M
[03/18 08:44:01 d2.utils.events]: eta: 5:57:33 iter: 2260 total_loss: 4.734 loss_box2d_reg: 0.3103 loss_box3d_depth: 1.641 loss_box3d_proj_ctr: 0.2636 loss_box3d_quat: 0.8372 loss_box3d_size: 0.1387 loss_centerness: 0.6398 loss_cls: 0.2915 loss_conf3d: 0.6569 lr: 0.002 max_mem: 5632M
[03/18 08:44:18 d2.utils.events]: eta: 5:30:56 iter: 2280 total_loss: 4.757 loss_box2d_reg: 0.3058 loss_box3d_depth: 1.65 loss_box3d_proj_ctr: 0.2666 loss_box3d_quat: 0.8246 loss_box3d_size: 0.1424 loss_centerness: 0.6401 loss_cls: 0.2887 loss_conf3d: 0.6513 lr: 0.002 max_mem: 5632M
[03/18 08:44:37 d2.utils.events]: eta: 5:56:43 iter: 2300 total_loss: 4.748 loss_box2d_reg: 0.2985 loss_box3d_depth: 1.617 loss_box3d_proj_ctr: 0.2756 loss_box3d_quat: 0.8398 loss_box3d_size: 0.1385 loss_centerness: 0.6378 loss_cls: 0.2865 loss_conf3d: 0.6559 lr: 0.002 max_mem: 5632M
[03/18 08:44:56 d2.utils.events]: eta: 6:03:35 iter: 2320 total_loss: 4.878 loss_box2d_reg: 0.3046 loss_box3d_depth: 1.738 loss_box3d_proj_ctr: 0.2641 loss_box3d_quat: 0.827 loss_box3d_size: 0.1369 loss_centerness: 0.6389 loss_cls: 0.2931 loss_conf3d: 0.6532 lr: 0.002 max_mem: 5632M
[03/18 08:45:15 d2.utils.events]: eta: 5:56:14 iter: 2340 total_loss: 4.75 loss_box2d_reg: 0.3079 loss_box3d_depth: 1.673 loss_box3d_proj_ctr: 0.2658 loss_box3d_quat: 0.8068 loss_box3d_size: 0.1362 loss_centerness: 0.6382 loss_cls: 0.2885 loss_conf3d: 0.6562 lr: 0.002 max_mem: 5632M
[03/18 08:45:34 d2.utils.events]: eta: 5:51:50 iter: 2360 total_loss: 4.779 loss_box2d_reg: 0.3051 loss_box3d_depth: 1.711 loss_box3d_proj_ctr: 0.2625 loss_box3d_quat: 0.7967 loss_box3d_size: 0.1338 loss_centerness: 0.6389 loss_cls: 0.287 loss_conf3d: 0.6534 lr: 0.002 max_mem: 5632M
[03/18 08:45:52 d2.utils.events]: eta: 5:46:12 iter: 2380 total_loss: 4.678 loss_box2d_reg: 0.3026 loss_box3d_depth: 1.561 loss_box3d_proj_ctr: 0.2674 loss_box3d_quat: 0.7952 loss_box3d_size: 0.14 loss_centerness: 0.6391 loss_cls: 0.2868 loss_conf3d: 0.6568 lr: 0.002 max_mem: 5632M
[03/18 08:46:10 d2.utils.events]: eta: 5:36:48 iter: 2400 total_loss: 4.632 loss_box2d_reg: 0.3105 loss_box3d_depth: 1.567 loss_box3d_proj_ctr: 0.2674 loss_box3d_quat: 0.7895 loss_box3d_size: 0.1395 loss_centerness: 0.64 loss_cls: 0.285 loss_conf3d: 0.6567 lr: 0.002 max_mem: 5632M
[03/18 08:46:29 d2.utils.events]: eta: 5:59:23 iter: 2420 total_loss: 4.619 loss_box2d_reg: 0.2976 loss_box3d_depth: 1.519 loss_box3d_proj_ctr: 0.256 loss_box3d_quat: 0.7626 loss_box3d_size: 0.1382 loss_centerness: 0.6371 loss_cls: 0.2782 loss_conf3d: 0.6571 lr: 0.002 max_mem: 5632M
[03/18 08:46:52 d2.utils.events]: eta: 7:09:39 iter: 2440 total_loss: 4.727 loss_box2d_reg: 0.3065 loss_box3d_depth: 1.608 loss_box3d_proj_ctr: 0.2692 loss_box3d_quat: 0.8576 loss_box3d_size: 0.142 loss_centerness: 0.6388 loss_cls: 0.282 loss_conf3d: 0.6544 lr: 0.002 max_mem: 5632M
[03/18 08:47:10 d2.utils.events]: eta: 5:42:01 iter: 2460 total_loss: 4.574 loss_box2d_reg: 0.2994 loss_box3d_depth: 1.496 loss_box3d_proj_ctr: 0.2655 loss_box3d_quat: 0.8225 loss_box3d_size: 0.1431 loss_centerness: 0.637 loss_cls: 0.275 loss_conf3d: 0.6546 lr: 0.002 max_mem: 5632M
[03/18 08:47:29 d2.utils.events]: eta: 5:55:06 iter: 2480 total_loss: 4.515 loss_box2d_reg: 0.2965 loss_box3d_depth: 1.429 loss_box3d_proj_ctr: 0.2601 loss_box3d_quat: 0.777 loss_box3d_size: 0.1336 loss_centerness: 0.6376 loss_cls: 0.2793 loss_conf3d: 0.6552 lr: 0.002 max_mem: 5632M
[03/18 08:47:48 d2.utils.events]: eta: 5:57:00 iter: 2500 total_loss: 4.605 loss_box2d_reg: 0.2945 loss_box3d_depth: 1.587 loss_box3d_proj_ctr: 0.2532 loss_box3d_quat: 0.7361 loss_box3d_size: 0.1298 loss_centerness: 0.6372 loss_cls: 0.2748 loss_conf3d: 0.6612 lr: 0.002 max_mem: 5632M
[03/18 08:48:06 d2.utils.events]: eta: 5:37:07 iter: 2520 total_loss: 4.575 loss_box2d_reg: 0.2975 loss_box3d_depth: 1.523 loss_box3d_proj_ctr: 0.2614 loss_box3d_quat: 0.784 loss_box3d_size: 0.1365 loss_centerness: 0.6379 loss_cls: 0.2823 loss_conf3d: 0.657 lr: 0.002 max_mem: 5632M
[03/18 08:48:25 d2.utils.events]: eta: 5:55:51 iter: 2540 total_loss: 4.446 loss_box2d_reg: 0.2961 loss_box3d_depth: 1.447 loss_box3d_proj_ctr: 0.2597 loss_box3d_quat: 0.7315 loss_box3d_size: 0.1346 loss_centerness: 0.6365 loss_cls: 0.2775 loss_conf3d: 0.6547 lr: 0.002 max_mem: 5632M
[03/18 08:48:44 d2.utils.events]: eta: 5:48:51 iter: 2560 total_loss: 4.474 loss_box2d_reg: 0.294 loss_box3d_depth: 1.449 loss_box3d_proj_ctr: 0.2583 loss_box3d_quat: 0.7227 loss_box3d_size: 0.135 loss_centerness: 0.6372 loss_cls: 0.2722 loss_conf3d: 0.6554 lr: 0.002 max_mem: 5632M
[03/18 08:49:01 d2.utils.events]: eta: 5:27:38 iter: 2580 total_loss: 4.447 loss_box2d_reg: 0.292 loss_box3d_depth: 1.434 loss_box3d_proj_ctr: 0.2508 loss_box3d_quat: 0.753 loss_box3d_size: 0.1375 loss_centerness: 0.6367 loss_cls: 0.2714 loss_conf3d: 0.657 lr: 0.002 max_mem: 5632M
[03/18 08:49:20 d2.utils.events]: eta: 5:54:22 iter: 2600 total_loss: 4.44 loss_box2d_reg: 0.2917 loss_box3d_depth: 1.458 loss_box3d_proj_ctr: 0.2665 loss_box3d_quat: 0.7522 loss_box3d_size: 0.1338 loss_centerness: 0.6376 loss_cls: 0.2746 loss_conf3d: 0.6521 lr: 0.002 max_mem: 5632M
[03/18 08:49:39 d2.utils.events]: eta: 5:41:48 iter: 2620 total_loss: 4.473 loss_box2d_reg: 0.2935 loss_box3d_depth: 1.505 loss_box3d_proj_ctr: 0.2562 loss_box3d_quat: 0.7129 loss_box3d_size: 0.136 loss_centerness: 0.6373 loss_cls: 0.2748 loss_conf3d: 0.6578 lr: 0.002 max_mem: 5632M
[03/18 08:49:56 d2.utils.events]: eta: 5:25:18 iter: 2640 total_loss: 4.442 loss_box2d_reg: 0.2907 loss_box3d_depth: 1.485 loss_box3d_proj_ctr: 0.2563 loss_box3d_quat: 0.761 loss_box3d_size: 0.1348 loss_centerness: 0.6371 loss_cls: 0.2694 loss_conf3d: 0.6534 lr: 0.002 max_mem: 5632M
[03/18 08:50:15 d2.utils.events]: eta: 5:49:01 iter: 2660 total_loss: 4.534 loss_box2d_reg: 0.2905 loss_box3d_depth: 1.566 loss_box3d_proj_ctr: 0.2603 loss_box3d_quat: 0.7051 loss_box3d_size: 0.1375 loss_centerness: 0.6372 loss_cls: 0.2721 loss_conf3d: 0.6562 lr: 0.002 max_mem: 5632M
[03/18 08:50:35 d2.utils.events]: eta: 6:05:16 iter: 2680 total_loss: 4.561 loss_box2d_reg: 0.2889 loss_box3d_depth: 1.559 loss_box3d_proj_ctr: 0.2582 loss_box3d_quat: 0.7043 loss_box3d_size: 0.1364 loss_centerness: 0.6367 loss_cls: 0.2717 loss_conf3d: 0.6565 lr: 0.002 max_mem: 5632M
[03/18 08:50:52 d2.utils.events]: eta: 5:31:49 iter: 2700 total_loss: 4.58 loss_box2d_reg: 0.292 loss_box3d_depth: 1.584 loss_box3d_proj_ctr: 0.258 loss_box3d_quat: 0.737 loss_box3d_size: 0.1324 loss_centerness: 0.6361 loss_cls: 0.2758 loss_conf3d: 0.6566 lr: 0.002 max_mem: 5632M
[03/18 08:51:11 d2.utils.events]: eta: 5:49:15 iter: 2720 total_loss: 4.496 loss_box2d_reg: 0.2858 loss_box3d_depth: 1.544 loss_box3d_proj_ctr: 0.2497 loss_box3d_quat: 0.7297 loss_box3d_size: 0.135 loss_centerness: 0.6359 loss_cls: 0.2638 loss_conf3d: 0.6561 lr: 0.002 max_mem: 5632M
[03/18 08:51:30 d2.utils.events]: eta: 5:41:28 iter: 2740 total_loss: 4.368 loss_box2d_reg: 0.2888 loss_box3d_depth: 1.386 loss_box3d_proj_ctr: 0.2586 loss_box3d_quat: 0.7447 loss_box3d_size: 0.133 loss_centerness: 0.6368 loss_cls: 0.2715 loss_conf3d: 0.6569 lr: 0.002 max_mem: 5632M
[03/18 08:51:47 d2.utils.events]: eta: 5:30:06 iter: 2760 total_loss: 4.388 loss_box2d_reg: 0.2878 loss_box3d_depth: 1.408 loss_box3d_proj_ctr: 0.2521 loss_box3d_quat: 0.7321 loss_box3d_size: 0.1348 loss_centerness: 0.636 loss_cls: 0.2677 loss_conf3d: 0.656 lr: 0.002 max_mem: 5632M
[03/18 08:52:07 d2.utils.events]: eta: 6:01:41 iter: 2780 total_loss: 4.387 loss_box2d_reg: 0.2832 loss_box3d_depth: 1.481 loss_box3d_proj_ctr: 0.2432 loss_box3d_quat: 0.7013 loss_box3d_size: 0.1309 loss_centerness: 0.6359 loss_cls: 0.2597 loss_conf3d: 0.6545 lr: 0.002 max_mem: 5632M
[03/18 08:52:26 d2.utils.events]: eta: 5:43:03 iter: 2800 total_loss: 4.456 loss_box2d_reg: 0.2914 loss_box3d_depth: 1.512 loss_box3d_proj_ctr: 0.2568 loss_box3d_quat: 0.7085 loss_box3d_size: 0.1354 loss_centerness: 0.6362 loss_cls: 0.2679 loss_conf3d: 0.6593 lr: 0.002 max_mem: 5632M
[03/18 08:52:44 d2.utils.events]: eta: 5:32:58 iter: 2820 total_loss: 4.351 loss_box2d_reg: 0.2867 loss_box3d_depth: 1.426 loss_box3d_proj_ctr: 0.2529 loss_box3d_quat: 0.6927 loss_box3d_size: 0.1331 loss_centerness: 0.6355 loss_cls: 0.2634 loss_conf3d: 0.6605 lr: 0.002 max_mem: 5632M
[03/18 08:53:03 d2.utils.events]: eta: 5:49:45 iter: 2840 total_loss: 4.384 loss_box2d_reg: 0.2846 loss_box3d_depth: 1.451 loss_box3d_proj_ctr: 0.2509 loss_box3d_quat: 0.6952 loss_box3d_size: 0.134 loss_centerness: 0.6362 loss_cls: 0.2656 loss_conf3d: 0.6546 lr: 0.002 max_mem: 5632M
[03/18 08:53:21 d2.utils.events]: eta: 5:42:01 iter: 2860 total_loss: 4.444 loss_box2d_reg: 0.284 loss_box3d_depth: 1.492 loss_box3d_proj_ctr: 0.2616 loss_box3d_quat: 0.6968 loss_box3d_size: 0.1322 loss_centerness: 0.6353 loss_cls: 0.2672 loss_conf3d: 0.6557 lr: 0.002 max_mem: 5632M
[03/18 08:53:40 d2.utils.events]: eta: 5:44:59 iter: 2880 total_loss: 4.309 loss_box2d_reg: 0.2822 loss_box3d_depth: 1.406 loss_box3d_proj_ctr: 0.2372 loss_box3d_quat: 0.6776 loss_box3d_size: 0.1315 loss_centerness: 0.6358 loss_cls: 0.2596 loss_conf3d: 0.6594 lr: 0.002 max_mem: 5632M
[03/18 08:53:58 d2.utils.events]: eta: 5:33:19 iter: 2900 total_loss: 4.426 loss_box2d_reg: 0.2839 loss_box3d_depth: 1.515 loss_box3d_proj_ctr: 0.2526 loss_box3d_quat: 0.6598 loss_box3d_size: 0.1345 loss_centerness: 0.6355 loss_cls: 0.2614 loss_conf3d: 0.6513 lr: 0.002 max_mem: 5632M
[03/18 08:54:17 d2.utils.events]: eta: 5:56:40 iter: 2920 total_loss: 4.361 loss_box2d_reg: 0.281 loss_box3d_depth: 1.49 loss_box3d_proj_ctr: 0.2437 loss_box3d_quat: 0.6613 loss_box3d_size: 0.1362 loss_centerness: 0.6353 loss_cls: 0.2619 loss_conf3d: 0.6571 lr: 0.002 max_mem: 5632M
[03/18 08:54:36 d2.utils.events]: eta: 5:44:03 iter: 2940 total_loss: 4.369 loss_box2d_reg: 0.2893 loss_box3d_depth: 1.403 loss_box3d_proj_ctr: 0.2484 loss_box3d_quat: 0.696 loss_box3d_size: 0.1297 loss_centerness: 0.6365 loss_cls: 0.2618 loss_conf3d: 0.658 lr: 0.002 max_mem: 5632M
[03/18 08:54:55 d2.utils.events]: eta: 5:43:49 iter: 2960 total_loss: 4.225 loss_box2d_reg: 0.2813 loss_box3d_depth: 1.339 loss_box3d_proj_ctr: 0.2536 loss_box3d_quat: 0.6704 loss_box3d_size: 0.1288 loss_centerness: 0.6345 loss_cls: 0.2615 loss_conf3d: 0.6535 lr: 0.002 max_mem: 5632M
[03/18 08:55:14 d2.utils.events]: eta: 5:45:25 iter: 2980 total_loss: 4.135 loss_box2d_reg: 0.2775 loss_box3d_depth: 1.281 loss_box3d_proj_ctr: 0.2419 loss_box3d_quat: 0.6571 loss_box3d_size: 0.1327 loss_centerness: 0.6345 loss_cls: 0.2553 loss_conf3d: 0.6543 lr: 0.002 max_mem: 5632M
[03/18 08:55:32 d2.utils.events]: eta: 5:43:44 iter: 3000 total_loss: 4.284 loss_box2d_reg: 0.2808 loss_box3d_depth: 1.398 loss_box3d_proj_ctr: 0.2512 loss_box3d_quat: 0.6717 loss_box3d_size: 0.132 loss_centerness: 0.6351 loss_cls: 0.255 loss_conf3d: 0.6505 lr: 0.002 max_mem: 5632M
[03/18 08:55:58 d2.utils.events]: eta: 7:43:59 iter: 3020 total_loss: 4.212 loss_box2d_reg: 0.2779 loss_box3d_depth: 1.328 loss_box3d_proj_ctr: 0.2439 loss_box3d_quat: 0.6287 loss_box3d_size: 0.1334 loss_centerness: 0.6352 loss_cls: 0.2549 loss_conf3d: 0.6532 lr: 0.002 max_mem: 5632M
[03/18 08:56:17 d2.utils.events]: eta: 5:51:41 iter: 3040 total_loss: 4.126 loss_box2d_reg: 0.2779 loss_box3d_depth: 1.247 loss_box3d_proj_ctr: 0.2383 loss_box3d_quat: 0.6291 loss_box3d_size: 0.1305 loss_centerness: 0.634 loss_cls: 0.2527 loss_conf3d: 0.653 lr: 0.002 max_mem: 5632M
[03/18 08:56:35 d2.utils.events]: eta: 5:31:32 iter: 3060 total_loss: 4.192 loss_box2d_reg: 0.2841 loss_box3d_depth: 1.366 loss_box3d_proj_ctr: 0.2451 loss_box3d_quat: 0.6337 loss_box3d_size: 0.1309 loss_centerness: 0.6356 loss_cls: 0.2591 loss_conf3d: 0.654 lr: 0.002 max_mem: 5632M
[03/18 08:56:53 d2.utils.events]: eta: 5:37:18 iter: 3080 total_loss: 4.522 loss_box2d_reg: 0.2818 loss_box3d_depth: 1.634 loss_box3d_proj_ctr: 0.2499 loss_box3d_quat: 0.6612 loss_box3d_size: 0.1304 loss_centerness: 0.6344 loss_cls: 0.2588 loss_conf3d: 0.6558 lr: 0.002 max_mem: 5632M
[03/18 08:57:12 d2.utils.events]: eta: 5:44:50 iter: 3100 total_loss: 4.261 loss_box2d_reg: 0.2773 loss_box3d_depth: 1.447 loss_box3d_proj_ctr: 0.2397 loss_box3d_quat: 0.6276 loss_box3d_size: 0.1336 loss_centerness: 0.6342 loss_cls: 0.255 loss_conf3d: 0.6566 lr: 0.002 max_mem: 5632M
[03/18 08:57:31 d2.utils.events]: eta: 5:37:49 iter: 3120 total_loss: 4.197 loss_box2d_reg: 0.276 loss_box3d_depth: 1.347 loss_box3d_proj_ctr: 0.2399 loss_box3d_quat: 0.6543 loss_box3d_size: 0.1347 loss_centerness: 0.6346 loss_cls: 0.2582 loss_conf3d: 0.6519 lr: 0.002 max_mem: 5632M
[03/18 08:57:50 d2.utils.events]: eta: 5:54:45 iter: 3140 total_loss: 4.451 loss_box2d_reg: 0.2801 loss_box3d_depth: 1.57 loss_box3d_proj_ctr: 0.2405 loss_box3d_quat: 0.6616 loss_box3d_size: 0.1349 loss_centerness: 0.6358 loss_cls: 0.2556 loss_conf3d: 0.6556 lr: 0.002 max_mem: 5632M
[03/18 08:58:10 d2.utils.events]: eta: 5:53:09 iter: 3160 total_loss: 4.202 loss_box2d_reg: 0.2782 loss_box3d_depth: 1.346 loss_box3d_proj_ctr: 0.2466 loss_box3d_quat: 0.6557 loss_box3d_size: 0.1314 loss_centerness: 0.6354 loss_cls: 0.2508 loss_conf3d: 0.662 lr: 0.002 max_mem: 5632M
[03/18 08:58:29 d2.utils.events]: eta: 5:43:45 iter: 3180 total_loss: 4.137 loss_box2d_reg: 0.2741 loss_box3d_depth: 1.32 loss_box3d_proj_ctr: 0.2406 loss_box3d_quat: 0.6324 loss_box3d_size: 0.1316 loss_centerness: 0.6332 loss_cls: 0.2508 loss_conf3d: 0.6543 lr: 0.002 max_mem: 5632M
[03/18 08:58:48 d2.utils.events]: eta: 5:52:24 iter: 3200 total_loss: 4.276 loss_box2d_reg: 0.2767 loss_box3d_depth: 1.485 loss_box3d_proj_ctr: 0.2449 loss_box3d_quat: 0.6627 loss_box3d_size: 0.129 loss_centerness: 0.6344 loss_cls: 0.2501 loss_conf3d: 0.6585 lr: 0.002 max_mem: 5632M
[03/18 08:59:07 d2.utils.events]: eta: 5:49:18 iter: 3220 total_loss: 4.165 loss_box2d_reg: 0.2779 loss_box3d_depth: 1.322 loss_box3d_proj_ctr: 0.2467 loss_box3d_quat: 0.6258 loss_box3d_size: 0.1308 loss_centerness: 0.6345 loss_cls: 0.2531 loss_conf3d: 0.6541 lr: 0.002 max_mem: 5632M
[03/18 08:59:26 d2.utils.events]: eta: 5:47:35 iter: 3240 total_loss: 4.029 loss_box2d_reg: 0.2724 loss_box3d_depth: 1.214 loss_box3d_proj_ctr: 0.2374 loss_box3d_quat: 0.6447 loss_box3d_size: 0.1283 loss_centerness: 0.6333 loss_cls: 0.2523 loss_conf3d: 0.6538 lr: 0.002 max_mem: 5632M
[03/18 08:59:45 d2.utils.events]: eta: 5:38:26 iter: 3260 total_loss: 4.382 loss_box2d_reg: 0.2703 loss_box3d_depth: 1.574 loss_box3d_proj_ctr: 0.2492 loss_box3d_quat: 0.6341 loss_box3d_size: 0.1268 loss_centerness: 0.6337 loss_cls: 0.248 loss_conf3d: 0.6577 lr: 0.002 max_mem: 5632M
[03/18 09:00:05 d2.utils.events]: eta: 5:51:17 iter: 3280 total_loss: 4.148 loss_box2d_reg: 0.2733 loss_box3d_depth: 1.327 loss_box3d_proj_ctr: 0.2404 loss_box3d_quat: 0.6175 loss_box3d_size: 0.1318 loss_centerness: 0.6339 loss_cls: 0.2458 loss_conf3d: 0.6578 lr: 0.002 max_mem: 5632M
[03/18 09:00:23 d2.utils.events]: eta: 5:26:54 iter: 3300 total_loss: 4.097 loss_box2d_reg: 0.2781 loss_box3d_depth: 1.297 loss_box3d_proj_ctr: 0.2451 loss_box3d_quat: 0.6235 loss_box3d_size: 0.1327 loss_centerness: 0.6343 loss_cls: 0.2471 loss_conf3d: 0.6524 lr: 0.002 max_mem: 5632M
[03/18 09:00:42 d2.utils.events]: eta: 5:43:12 iter: 3320 total_loss: 4.033 loss_box2d_reg: 0.2727 loss_box3d_depth: 1.255 loss_box3d_proj_ctr: 0.2418 loss_box3d_quat: 0.5899 loss_box3d_size: 0.1295 loss_centerness: 0.6344 loss_cls: 0.2488 loss_conf3d: 0.6543 lr: 0.002 max_mem: 5632M
[03/18 09:01:01 d2.utils.events]: eta: 5:49:41 iter: 3340 total_loss: 3.927 loss_box2d_reg: 0.2716 loss_box3d_depth: 1.178 loss_box3d_proj_ctr: 0.2358 loss_box3d_quat: 0.5726 loss_box3d_size: 0.1281 loss_centerness: 0.6328 loss_cls: 0.2438 loss_conf3d: 0.6518 lr: 0.002 max_mem: 5632M
[03/18 09:01:20 d2.utils.events]: eta: 5:33:51 iter: 3360 total_loss: 4.069 loss_box2d_reg: 0.2667 loss_box3d_depth: 1.303 loss_box3d_proj_ctr: 0.2352 loss_box3d_quat: 0.598 loss_box3d_size: 0.1264 loss_centerness: 0.6333 loss_cls: 0.244 loss_conf3d: 0.6528 lr: 0.002 max_mem: 5632M
[03/18 09:01:39 d2.utils.events]: eta: 5:43:53 iter: 3380 total_loss: 4.315 loss_box2d_reg: 0.2697 loss_box3d_depth: 1.498 loss_box3d_proj_ctr: 0.2312 loss_box3d_quat: 0.6348 loss_box3d_size: 0.1275 loss_centerness: 0.6341 loss_cls: 0.2473 loss_conf3d: 0.6571 lr: 0.002 max_mem: 5632M
[03/18 09:01:59 d2.utils.events]: eta: 5:59:36 iter: 3400 total_loss: 4.148 loss_box2d_reg: 0.2712 loss_box3d_depth: 1.356 loss_box3d_proj_ctr: 0.2322 loss_box3d_quat: 0.6197 loss_box3d_size: 0.1281 loss_centerness: 0.6334 loss_cls: 0.2436 loss_conf3d: 0.6569 lr: 0.002 max_mem: 5632M
[03/18 09:02:17 d2.utils.events]: eta: 5:35:54 iter: 3420 total_loss: 4.261 loss_box2d_reg: 0.2712 loss_box3d_depth: 1.412 loss_box3d_proj_ctr: 0.2381 loss_box3d_quat: 0.645 loss_box3d_size: 0.1313 loss_centerness: 0.6326 loss_cls: 0.2482 loss_conf3d: 0.6547 lr: 0.002 max_mem: 5632M
[03/18 09:02:36 d2.utils.events]: eta: 5:45:17 iter: 3440 total_loss: 4.218 loss_box2d_reg: 0.2671 loss_box3d_depth: 1.456 loss_box3d_proj_ctr: 0.2377 loss_box3d_quat: 0.6277 loss_box3d_size: 0.13 loss_centerness: 0.6329 loss_cls: 0.2465 loss_conf3d: 0.6572 lr: 0.002 max_mem: 5632M
[03/18 09:02:56 d2.utils.events]: eta: 5:54:58 iter: 3460 total_loss: 4.117 loss_box2d_reg: 0.2729 loss_box3d_depth: 1.319 loss_box3d_proj_ctr: 0.2372 loss_box3d_quat: 0.5696 loss_box3d_size: 0.1299 loss_centerness: 0.6342 loss_cls: 0.2447 loss_conf3d: 0.6547 lr: 0.002 max_mem: 5632M
[03/18 09:03:15 d2.utils.events]: eta: 5:33:40 iter: 3480 total_loss: 3.874 loss_box2d_reg: 0.2655 loss_box3d_depth: 1.147 loss_box3d_proj_ctr: 0.2357 loss_box3d_quat: 0.5902 loss_box3d_size: 0.1294 loss_centerness: 0.6319 loss_cls: 0.2414 loss_conf3d: 0.6477 lr: 0.002 max_mem: 5632M
[03/18 09:03:34 d2.utils.events]: eta: 5:45:55 iter: 3500 total_loss: 3.904 loss_box2d_reg: 0.2666 loss_box3d_depth: 1.196 loss_box3d_proj_ctr: 0.2338 loss_box3d_quat: 0.5882 loss_box3d_size: 0.1243 loss_centerness: 0.6319 loss_cls: 0.2413 loss_conf3d: 0.6465 lr: 0.002 max_mem: 5632M
[03/18 09:03:53 d2.utils.events]: eta: 5:37:07 iter: 3520 total_loss: 3.903 loss_box2d_reg: 0.2662 loss_box3d_depth: 1.189 loss_box3d_proj_ctr: 0.2242 loss_box3d_quat: 0.5488 loss_box3d_size: 0.1269 loss_centerness: 0.6321 loss_cls: 0.2358 loss_conf3d: 0.6463 lr: 0.002 max_mem: 5632M
[03/18 09:04:11 d2.utils.events]: eta: 5:27:48 iter: 3540 total_loss: 3.991 loss_box2d_reg: 0.2662 loss_box3d_depth: 1.226 loss_box3d_proj_ctr: 0.2323 loss_box3d_quat: 0.5794 loss_box3d_size: 0.1261 loss_centerness: 0.6322 loss_cls: 0.2415 loss_conf3d: 0.6599 lr: 0.002 max_mem: 5632M
[03/18 09:04:31 d2.utils.events]: eta: 5:43:34 iter: 3560 total_loss: 3.923 loss_box2d_reg: 0.2616 loss_box3d_depth: 1.201 loss_box3d_proj_ctr: 0.2304 loss_box3d_quat: 0.5719 loss_box3d_size: 0.1247 loss_centerness: 0.6322 loss_cls: 0.2418 loss_conf3d: 0.6518 lr: 0.002 max_mem: 5632M
[03/18 09:04:51 d2.utils.events]: eta: 5:58:03 iter: 3580 total_loss: 3.973 loss_box2d_reg: 0.2671 loss_box3d_depth: 1.211 loss_box3d_proj_ctr: 0.2281 loss_box3d_quat: 0.5739 loss_box3d_size: 0.1277 loss_centerness: 0.6316 loss_cls: 0.2396 loss_conf3d: 0.6473 lr: 0.002 max_mem: 5632M
[03/18 09:05:12 d2.utils.events]: eta: 6:27:45 iter: 3600 total_loss: 3.881 loss_box2d_reg: 0.2625 loss_box3d_depth: 1.146 loss_box3d_proj_ctr: 0.2291 loss_box3d_quat: 0.5851 loss_box3d_size: 0.1251 loss_centerness: 0.6316 loss_cls: 0.2373 loss_conf3d: 0.6507 lr: 0.002 max_mem: 5632M
[03/18 09:05:33 d2.utils.events]: eta: 6:12:17 iter: 3620 total_loss: 3.976 loss_box2d_reg: 0.2611 loss_box3d_depth: 1.196 loss_box3d_proj_ctr: 0.2287 loss_box3d_quat: 0.6027 loss_box3d_size: 0.124 loss_centerness: 0.632 loss_cls: 0.242 loss_conf3d: 0.6493 lr: 0.002 max_mem: 5632M
[03/18 09:05:53 d2.utils.events]: eta: 5:43:25 iter: 3640 total_loss: 3.925 loss_box2d_reg: 0.2619 loss_box3d_depth: 1.216 loss_box3d_proj_ctr: 0.2267 loss_box3d_quat: 0.5706 loss_box3d_size: 0.1264 loss_centerness: 0.6323 loss_cls: 0.2392 loss_conf3d: 0.6489 lr: 0.002 max_mem: 5632M
[03/18 09:06:11 d2.utils.events]: eta: 5:19:53 iter: 3660 total_loss: 4.046 loss_box2d_reg: 0.2622 loss_box3d_depth: 1.28 loss_box3d_proj_ctr: 0.2336 loss_box3d_quat: 0.5887 loss_box3d_size: 0.1281 loss_centerness: 0.632 loss_cls: 0.2395 loss_conf3d: 0.6507 lr: 0.002 max_mem: 5632M
[03/18 09:06:29 d2.utils.events]: eta: 5:31:13 iter: 3680 total_loss: 3.929 loss_box2d_reg: 0.2644 loss_box3d_depth: 1.235 loss_box3d_proj_ctr: 0.2333 loss_box3d_quat: 0.5652 loss_box3d_size: 0.1252 loss_centerness: 0.6313 loss_cls: 0.2365 loss_conf3d: 0.65 lr: 0.002 max_mem: 5632M
[03/18 09:06:48 d2.utils.events]: eta: 5:35:35 iter: 3700 total_loss: 3.76 loss_box2d_reg: 0.2559 loss_box3d_depth: 1.153 loss_box3d_proj_ctr: 0.2246 loss_box3d_quat: 0.5464 loss_box3d_size: 0.1246 loss_centerness: 0.6309 loss_cls: 0.2323 loss_conf3d: 0.6416 lr: 0.002 max_mem: 5632M
[03/18 09:07:08 d2.utils.events]: eta: 5:50:52 iter: 3720 total_loss: 3.931 loss_box2d_reg: 0.2599 loss_box3d_depth: 1.221 loss_box3d_proj_ctr: 0.2316 loss_box3d_quat: 0.5674 loss_box3d_size: 0.1215 loss_centerness: 0.6313 loss_cls: 0.2366 loss_conf3d: 0.654 lr: 0.002 max_mem: 5632M
[03/18 09:07:30 d2.utils.events]: eta: 6:38:46 iter: 3740 total_loss: 3.85 loss_box2d_reg: 0.2601 loss_box3d_depth: 1.181 loss_box3d_proj_ctr: 0.2173 loss_box3d_quat: 0.5407 loss_box3d_size: 0.1272 loss_centerness: 0.6313 loss_cls: 0.2356 loss_conf3d: 0.6511 lr: 0.002 max_mem: 5632M
[03/18 09:07:51 d2.utils.events]: eta: 6:05:52 iter: 3760 total_loss: 3.878 loss_box2d_reg: 0.2626 loss_box3d_depth: 1.163 loss_box3d_proj_ctr: 0.226 loss_box3d_quat: 0.5651 loss_box3d_size: 0.1229 loss_centerness: 0.6318 loss_cls: 0.2315 loss_conf3d: 0.6493 lr: 0.002 max_mem: 5632M
[03/18 09:08:10 d2.utils.events]: eta: 5:34:07 iter: 3780 total_loss: 3.955 loss_box2d_reg: 0.2578 loss_box3d_depth: 1.291 loss_box3d_proj_ctr: 0.2254 loss_box3d_quat: 0.5275 loss_box3d_size: 0.1226 loss_centerness: 0.6312 loss_cls: 0.2327 loss_conf3d: 0.6544 lr: 0.002 max_mem: 5632M
[03/18 09:08:29 d2.utils.events]: eta: 5:28:27 iter: 3800 total_loss: 3.901 loss_box2d_reg: 0.2607 loss_box3d_depth: 1.223 loss_box3d_proj_ctr: 0.2275 loss_box3d_quat: 0.5669 loss_box3d_size: 0.1246 loss_centerness: 0.632 loss_cls: 0.2323 loss_conf3d: 0.6478 lr: 0.002 max_mem: 5632M
[03/18 09:08:48 d2.utils.events]: eta: 5:41:50 iter: 3820 total_loss: 3.867 loss_box2d_reg: 0.2533 loss_box3d_depth: 1.197 loss_box3d_proj_ctr: 0.2283 loss_box3d_quat: 0.5565 loss_box3d_size: 0.1233 loss_centerness: 0.6303 loss_cls: 0.2357 loss_conf3d: 0.6465 lr: 0.002 max_mem: 5632M
[03/18 09:09:05 d2.utils.events]: eta: 5:07:06 iter: 3840 total_loss: 3.88 loss_box2d_reg: 0.2575 loss_box3d_depth: 1.137 loss_box3d_proj_ctr: 0.2199 loss_box3d_quat: 0.5358 loss_box3d_size: 0.1231 loss_centerness: 0.6316 loss_cls: 0.2371 loss_conf3d: 0.6506 lr: 0.002 max_mem: 5632M
[03/18 09:09:24 d2.utils.events]: eta: 5:35:23 iter: 3860 total_loss: 3.827 loss_box2d_reg: 0.2545 loss_box3d_depth: 1.159 loss_box3d_proj_ctr: 0.2218 loss_box3d_quat: 0.5643 loss_box3d_size: 0.1228 loss_centerness: 0.6317 loss_cls: 0.2303 loss_conf3d: 0.6469 lr: 0.002 max_mem: 5632M
[03/18 09:09:44 d2.utils.events]: eta: 5:37:09 iter: 3880 total_loss: 3.797 loss_box2d_reg: 0.2574 loss_box3d_depth: 1.102 loss_box3d_proj_ctr: 0.2225 loss_box3d_quat: 0.5656 loss_box3d_size: 0.1252 loss_centerness: 0.6315 loss_cls: 0.2341 loss_conf3d: 0.6412 lr: 0.002 max_mem: 5632M
[03/18 09:10:01 d2.utils.events]: eta: 5:09:14 iter: 3900 total_loss: 3.722 loss_box2d_reg: 0.257 loss_box3d_depth: 1.062 loss_box3d_proj_ctr: 0.2175 loss_box3d_quat: 0.5243 loss_box3d_size: 0.1242 loss_centerness: 0.6314 loss_cls: 0.2312 loss_conf3d: 0.6467 lr: 0.002 max_mem: 5632M
[03/18 09:10:21 d2.utils.events]: eta: 5:43:19 iter: 3920 total_loss: 3.694 loss_box2d_reg: 0.2505 loss_box3d_depth: 1.063 loss_box3d_proj_ctr: 0.2193 loss_box3d_quat: 0.5215 loss_box3d_size: 0.1188 loss_centerness: 0.6305 loss_cls: 0.2219 loss_conf3d: 0.6432 lr: 0.002 max_mem: 5632M
[03/18 09:10:40 d2.utils.events]: eta: 5:38:46 iter: 3940 total_loss: 3.959 loss_box2d_reg: 0.2537 loss_box3d_depth: 1.285 loss_box3d_proj_ctr: 0.228 loss_box3d_quat: 0.5264 loss_box3d_size: 0.1255 loss_centerness: 0.6307 loss_cls: 0.2279 loss_conf3d: 0.6532 lr: 0.002 max_mem: 5632M
[03/18 09:11:00 d2.utils.events]: eta: 5:56:42 iter: 3960 total_loss: 3.935 loss_box2d_reg: 0.2608 loss_box3d_depth: 1.202 loss_box3d_proj_ctr: 0.2238 loss_box3d_quat: 0.5946 loss_box3d_size: 0.1248 loss_centerness: 0.6324 loss_cls: 0.2327 loss_conf3d: 0.647 lr: 0.002 max_mem: 5632M
[03/18 09:11:19 d2.utils.events]: eta: 5:27:30 iter: 3980 total_loss: 3.936 loss_box2d_reg: 0.2561 loss_box3d_depth: 1.241 loss_box3d_proj_ctr: 0.228 loss_box3d_quat: 0.606 loss_box3d_size: 0.1266 loss_centerness: 0.6312 loss_cls: 0.2326 loss_conf3d: 0.6486 lr: 0.002 max_mem: 5632M
[03/18 09:11:38 d2.utils.events]: eta: 5:37:11 iter: 4000 total_loss: 3.912 loss_box2d_reg: 0.2555 loss_box3d_depth: 1.18 loss_box3d_proj_ctr: 0.2265 loss_box3d_quat: 0.5709 loss_box3d_size: 0.1198 loss_centerness: 0.63 loss_cls: 0.23 loss_conf3d: 0.6514 lr: 0.002 max_mem: 5632M
[03/18 09:11:44 fvcore.common.checkpoint]: Saving checkpoint to ./model_0003999.pth
[03/18 09:11:45 tridet.data.dataset_mappers.dataset_mapper]: Augmentations used in training: [ResizeShortestEdge(short_edge_length=(384, 384), max_size=100000, sample_style='choice')]
[03/18 09:11:45 d2.data.common]: Serializing 3769 elements to byte tensors and concatenating them all ...
[03/18 09:11:46 d2.data.common]: Serialized dataset takes 13.38 MiB
[03/18 09:11:46 tridet.data.build]: Using test sampler InferenceSampler
[03/18 09:11:46 d2.evaluation.evaluator]: Start inference on 48 batches

question about allocentric_to_egocentric

Hi, actually i have some problem with this (allocentric_to_egocentric).
I can't really understand how 'R_local_to_global' come from after i read code and paper.
Could you help recommend some knowledge material about this question? I would really appreciate that

I think it's something like 'alpha' observation angle in KITTI.(To get rot_y, we need to consider 'alpha' and ray angle.)
But i can't exact reason this process in DD3D. (allocentric_to_egocentric)

question about depth pretraining

Could you please provide more details or share the tools about the generation of depth maps for pretraining?

The "remapped" for the pretrained model ?

Hello, what is the "remapped" mean in the pretrained model (https://tri-ml-public.s3.amazonaws.com/github/dd3d/pretrained/depth_pretrained_dla34-y1urdmir-20210422_165446-model_final-remapped.pth)? If I want to train DD3D using DLA34 (only pretrained on Imagenet), what transformation need to do for DLA model parameters ?

Export ONNX model

Hello

Could you share your onnx export script, which can help us a lot.

Thanks!

Can I get the Eigen-clean split?

Hi,
Can I get the Eigen-clean split mentioned in the paper?

pos_inds.numel() == 0 is not allowed in fact

in fcos3d.py: FCOS3DLoss, it writes

if pos_inds.numel() == 0:
            losses = {
                "loss_box3d_quat": box3d_quat.sum() * 0.,
                "loss_box3d_proj_ctr": box3d_ctr.sum() * 0.,
                "loss_box3d_depth": box3d_depth.sum() * 0.,
                "loss_box3d_size": box3d_size.sum() * 0.,
                "loss_conf3d": box3d_conf.sum() * 0.
            }
            return losses

        if len(labels) != len(box3d_targets):
            raise ValueError(
                f"The size of 'labels' and 'box3d_targets' does not match: a={len(labels)}, b={len(box3d_targets)}"
            )

        num_classes = self.num_classes if not self.class_agnostic else 1

        box3d_quat_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, 4, num_classes) for x in box3d_quat])
        box3d_ctr_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, 2, num_classes) for x in box3d_ctr])
        box3d_depth_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, num_classes) for x in box3d_depth])
        box3d_size_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, 3, num_classes) for x in box3d_size])
        box3d_conf_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, num_classes) for x in box3d_conf])

but it should be fixed as

if len(labels) != len(box3d_targets):
            raise ValueError(
                f"The size of 'labels' and 'box3d_targets' does not match: a={len(labels)}, b={len(box3d_targets)}"
            )

        num_classes = self.num_classes if not self.class_agnostic else 1

        box3d_quat_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, 4, num_classes) for x in box3d_quat])
        box3d_ctr_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, 2, num_classes) for x in box3d_ctr])
        box3d_depth_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, num_classes) for x in box3d_depth])
        box3d_size_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, 3, num_classes) for x in box3d_size])
        box3d_conf_pred = cat([x.permute(0, 2, 3, 1).reshape(-1, num_classes) for x in box3d_conf])

        ## ori author got wrong here, they put it above cut
        if pos_inds.numel() == 0:
            losses = {
                "loss_box3d_quat": box3d_quat_pred.sum() * 0.,
                "loss_box3d_proj_ctr": box3d_ctr_pred.sum() * 0.,
                "loss_box3d_depth": box3d_depth_pred.sum() * 0.,
                "loss_box3d_size": box3d_size_pred.sum() * 0.,
                "loss_conf3d": box3d_conf_pred.sum() * 0.
            }
            return losses

other wise, .sum() is not allowed

in fcos2d.py: FCOS2DLoss, it write:

if pos_inds.numel() == 0:
            losses = {
                "loss_cls": loss_cls,
                "loss_box2d_reg": box2d_reg_pred.sum() * 0.,
                "loss_centerness": centerness_pred.sum() * 0.,
            }
            return losses, {}

        # NOTE: The rest of losses only consider foreground pixels.
        box2d_reg_pred = box2d_reg_pred[pos_inds]
        box2d_reg_targets = box2d_reg_targets[pos_inds]

        centerness_pred = centerness_pred[pos_inds]

        # Compute centerness targets here using 2D regression targets of foreground pixels.
        centerness_targets = compute_ctrness_targets(box2d_reg_targets)

        # Denominator for all foreground losses.
        ctrness_targets_sum = centerness_targets.sum()
        loss_denom = max(reduce_sum(ctrness_targets_sum).item() / num_gpus, 1e-6)

but the fact is, if one card returns (pos_inds.numel()==0), the other card would be waiting in reduce_sum. so the code is just waiting and stucks on cuda's reduction .
finally, it was modified like this:

# used for reduce_sum(ctrness_targets_sum) below, as all reduce | reduce_sum is used !!!!
        ctrness_targets_sum = centerness_pred.sum() * 0. 
        _loss_box2d_reg = box2d_reg_pred.sum() * 0.
        loss_centerness = centerness_pred.sum() * 0.
        if pos_inds.numel() != 0:
            # NOTE: The rest of losses only consider foreground pixels.
            box2d_reg_pred = box2d_reg_pred[pos_inds]
            box2d_reg_targets = box2d_reg_targets[pos_inds]
            centerness_pred = centerness_pred[pos_inds]
        
            # Compute centerness targets here using 2D regression targets of foreground pixels.
            centerness_targets = compute_ctrness_targets(box2d_reg_targets)

            # Denominator for all foreground losses.
            ctrness_targets_sum = centerness_targets.sum()

            # ----------------------
            # 2D box regression loss
            # ----------------------
            _loss_box2d_reg = self.box2d_reg_loss_fn(box2d_reg_pred, box2d_reg_targets, centerness_targets)

            # ---------------
            # Centerness loss
            # ---------------
            loss_centerness = F.binary_cross_entropy_with_logits(
                centerness_pred, centerness_targets, reduction="sum"
            ) / num_pos_avg
        # else:
            # print(".ret", end=" ", flush=True)

        # print(".red", end=" ", flush=True)
        loss_denom = max(reduce_sum(ctrness_targets_sum).item() / num_gpus, 1e-6)
        loss_box2d_reg = _loss_box2d_reg / loss_denom

        loss_dict = {"loss_cls": loss_cls, "loss_box2d_reg": loss_box2d_reg, "loss_centerness": loss_centerness}
        extra_info = {"loss_denom": loss_denom, "centerness_targets": centerness_targets} if pos_inds.numel() != 0 else {}

        # print(".f", end=" ", flush=True)
        return loss_dict, extra_info

Dense Depth Config file

Hello @dennis-park-TRI , thank you for your work in DD3D! That's very insightful!

As I go through the repo, I can find some code related to depth pertaining like DenseDepthHead and so on. However, I didn't find such a config to start the depth pretraining. Could you provide your config file for pretraining depth on the V99 model? I assume the data format can be in KITTI since it provides Velodyne data.

Again, thank you for opensource this awesome repo!

Classes Pedestrian and Cyclist much lower than paper

I uesd released weight on GitHub, KITTI DLA-34 and KITTI V2-99.
And I tried to use both weight to do evaluation on validation set. I could get similer results on Car class, but on other classes are all limited to zero.
I known the TTA and different evaluation sets are cause variance. But the variance should not be so much. Are there any missing settings that are causing this issue?
Below table is my evaluate results:

Car AP [email protected]		Paper	KITTI Submit	KITTI DLA	KITTI V2
BEV AP	Easy	30.98	32.35	31.65	40.70
	Med	22.56	23.41	24.43	32.04
	Hard	20.03	20.42	21.72	28.54
3D AP	Easy	23.22	23.19	22.56	30.38
	Med	16.34	16.87	16.98	23.73
	Hard	14.20	14.36	14.93	20.88

Pedestrian AP [email protected]		Paper	KITTI Submit	KITTI DLA	KITTI V2
BEV AP	Easy	15.90	18.58	0.04	0.538
	Med	10.85	12.51	0.03	0.056
	Hard	8.05	10.65	0.01	0.028
3D AP	Easy	13.91	16.64	0.007	0.017
	Med	9.30	11.04	0.008	0.018
	Hard	8.05	9.38	0.009	0.019

Cyclist AP [email protected]		Paper	KITTI Submit	KITTI DLA	KITTI V2
BEV AP	Easy	3.20	9.20	0.44	0.447
	Med	1.99	5.69	0.27	0.229
	Hard	1.79	5.20	0.28	0.258
3D AP	Easy	2.39	7.52	0.13	0.126
	Med	1.52	4.79	0.12	0.123
	Hard	1.31	4.22	0.11	0.111

And I have another question about published results, why it has differece between arxiv paper and kitti submit.

Thank you so much.

How to pretrain depth using a different depth dataset

How can we pretrain depth using a different/custom depth dataset?
There is a dense depth head in the provided code, but could you please provide instructions on how to use in terms of how to attach it to the model and whether we have to only consider dense depth loss while training it or the total loss?
An example would really help :)

Thank You

where can i get the DDAD15M dataset?

HELLO, THANKS FOR THE GOOD JOB.
And i wonder where can i get the DDAD15M dataset? I can only find DDAD dataset on github.
THANKS.

Box pose transfomation from camera 2 to velodyne may not be correct.

Hi, I found code at

dd3d/tridet/data/datasets/kitti_3d/build.py

Line 260 in 86d8660

box_pose = pose_0S * pose_02.inverse() * box_pose

may not be correct.
If we want to get the box pose in the velodyne frame, I think the code may be box_pose = pose_0V.inverse() * pose_02 * box_pose.
I am not sure if it is correct, so I hope you can double check, thank you!

Inference Example

Could you please provide a simple inference example?

Failed to fetch https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/Packages.gz

lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 18.04.6 LTS
Release: 18.04
Codename: bionic

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 1 542 G /usr/lib/xorg/Xorg 583MiB |
+-----------------------------------------------------------------------------+

---> Running in 48fca625b22f
Get:1 http://archive.ubuntu.com/ubuntu bionic InRelease [242 kB]
Get:2 http://security.ubuntu.com/ubuntu bionic-security InRelease [88.7 kB]
Get:3 http://archive.ubuntu.com/ubuntu bionic-updates InRelease [88.7 kB]
Get:4 http://archive.ubuntu.com/ubuntu bionic-backports InRelease [74.6 kB]
Get:5 http://security.ubuntu.com/ubuntu bionic-security/universe amd64 Packages [1486 kB]
Ign:6 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 InRelease
Get:7 http://security.ubuntu.com/ubuntu bionic-security/main amd64 Packages [2684 kB]
Get:8 http://archive.ubuntu.com/ubuntu bionic/multiverse amd64 Packages [186 kB]
Get:9 http://security.ubuntu.com/ubuntu bionic-security/restricted amd64 Packages [879 kB]
Get:10 http://archive.ubuntu.com/ubuntu bionic/main amd64 Packages [1344 kB]
Get:11 http://security.ubuntu.com/ubuntu bionic-security/multiverse amd64 Packages [21.1 kB]
Get:12 http://archive.ubuntu.com/ubuntu bionic/universe amd64 Packages [11.3 MB]
Ign:13 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64 InRelease
Get:14 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 Release [696 B]
Get:15 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64 Release [564 B]
Get:16 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 Release.gpg [836 B]
Get:17 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64 Release.gpg [833 B]
Get:18 http://archive.ubuntu.com/ubuntu bionic/restricted amd64 Packages [13.5 kB]
Get:19 http://archive.ubuntu.com/ubuntu bionic-updates/main amd64 Packages [3129 kB]
Get:20 http://archive.ubuntu.com/ubuntu bionic-updates/restricted amd64 Packages [916 kB]
Get:21 http://archive.ubuntu.com/ubuntu bionic-updates/universe amd64 Packages [2264 kB]
Get:22 http://archive.ubuntu.com/ubuntu bionic-updates/multiverse amd64 Packages [29.8 kB]
Get:23 http://archive.ubuntu.com/ubuntu bionic-backports/universe amd64 Packages [12.9 kB]
Get:24 http://archive.ubuntu.com/ubuntu bionic-backports/main amd64 Packages [12.2 kB]
Get:25 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 Packages [950 kB]
Err:25 https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 Packages
File has unexpected size (945744 != 949929). Mirror sync in progress? [IP: 152.195.19.142 443]
Hashes of expected file:

Filesize:949929 [weak]
SHA256:b1ce21ca45ffb1607383f6be3d60d830af7d5a1ccb7fd8863740338a5eda8999
SHA1:238dac26b27ef48ce0f00924699a64edad6a367f [weak]
MD5Sum:f497438f74da540f89ac169f32fc926f [weak]
Release file created at: Wed, 30 Mar 2022 15:01:50 +0000
Get:26 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64 Packages [73.8 kB]
Fetched 24.9 MB in 2s (13.3 MB/s)
Reading package lists...
E: Failed to fetch https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/Packages.gz File has unexpected size (945744 != 949929). Mirror sync in progress? [IP: 152.195.19.142 443]
Hashes of expected file:
- Filesize:949929 [weak]
- SHA256:b1ce21ca45ffb1607383f6be3d60d830af7d5a1ccb7fd8863740338a5eda8999
- SHA1:238dac26b27ef48ce0f00924699a64edad6a367f [weak]
- MD5Sum:f497438f74da540f89ac169f32fc926f [weak]
Release file created at: Wed, 30 Mar 2022 15:01:50 +0000
E: Some index files failed to download. They have been ignored, or old ones used instead.
The command '/bin/sh -c apt-get update && apt-get install -y build-essential cmake ffmpeg g++-4.8 git curl docker.io vim wget unzip htop libjpeg-dev libpng-dev libavdevice-dev pkg-config python${PYTHON_VERSION} python${PYTHON_VERSION}-dev python3-tk python${PYTHON_VERSION}-distutils python3-opencv && ln -sf /usr/bin/python${PYTHON_VERSION} /usr/bin/python && ln -sf /usr/bin/python${PYTHON_VERSION} /usr/bin/python3 && rm -rf /var/lib/apt/lists/*' returned a non-zero code: 100
Makefile:49: recipe for target 'docker-build' failed
make: *** [docker-build] Error 100

I can't reproduce the result.

Hi,
The backbone is DLA34. The GPU number is 2 and per-GPU bachsize is 2.The result is below.

100%|█████████████████████████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00,  6.42it/s]
[12/30 19:06:08 tridet]: Running prediction visualizer: box3d_visualizer
100%|█████████████████████████████████████████████████████████████████████████████████████| 20/20 [00:12<00:00,  1.65it/s]
[12/30 19:06:27 tridet]: Evaluation results for kitti_3d_val in csv format:
[12/30 19:06:27 tridet.utils.train]: Test results:
| metric                                                   | value     |
|:---------------------------------------------------------|:----------|
| kitti_3d_val/bbox-tta/AP                                 | 1.17012   |
| kitti_3d_val/bbox-tta/AP50                               | 3.13831   |
| kitti_3d_val/bbox-tta/AP75                               | 0.587498  |
| kitti_3d_val/bbox-tta/APs                                | 0         |
| kitti_3d_val/bbox-tta/APm                                | 1.25059   |
| kitti_3d_val/bbox-tta/APl                                | 2.15315   |
| kitti_3d_val/bbox-tta/AP-Car                             | 5.8506    |
| kitti_3d_val/bbox-tta/AP-Pedestrian                      | 0         |
| kitti_3d_val/bbox-tta/AP-Cyclist                         | 0         |
| kitti_3d_val/bbox-tta/AP-Van                             | 0         |
| kitti_3d_val/bbox-tta/AP-Truck                           | 0         |
| kitti_3d_val/kitti_box3d_r40/Car_Easy_0.5-tta            | 2.79858   |
| kitti_3d_val/kitti_box3d_r40/Car_Easy_0.7-tta            | 0         |
| kitti_3d_val/kitti_box3d_r40/Car_Moderate_0.5-tta        | 1.21186   |
| kitti_3d_val/kitti_box3d_r40/Car_Moderate_0.7-tta        | 0         |
| kitti_3d_val/kitti_box3d_r40/Car_Hard_0.5-tta            | 0.796626  |
| kitti_3d_val/kitti_box3d_r40/Car_Hard_0.7-tta            | 0         |
```

How to calc 'MEAN_DEPTH_PER_LEVEL' and 'STD_DEPTH_PER_LEVEL'

@dennis-park-TRI Hi, I have some question about mean_depth_per_level and std_depth_per_level in config/train_datasets/kitti_3d.yaml。how to calc them,can you provide some supports for me ?Many thanks...

Detail for the depth pre-training's fcos-head

Could you please provide more detail for the depth pre-training?
I want to know the difference between pre-training and training's focs-head!
thanks a lot, it will be very helpful!

why dd3d_kitti_omninets.yaml feature_extractors use DLA34

https://github.com/TRI-ML/dd3d/blob/main/configs/experiments/dd3d_kitti_omninets.yaml#L7 override /feature_extractors@FE: dla34_fpn

How to training two datasets with different camera intrinsic together

Hi,

Thank you very much for release source code!

I have three basic questions

Can I train two datasets (with different camera intrinsic) in one mini-batch by just pass different intrinsic param to loss function?
In the paper, you mentioned the image could be resized during training, in which situation we need to do the image resize except the pre-train?
Is the V2-99 backbone model able to use in the real-time(more than 25 fps) scenario?

Thanks

Running inference with visualization

Hi there, I'm trying to test out the pre-trained model on a custom dataset, but I am unable to find the script that does that (Seems like visualize_dataloader is visualizing the GT). I've been trying to understand the hydra experiments but I am not sure if there is an option to do just prediction without ground truths.

Any help on this would be highly appreciated, thanks!

ps. this is assuming I have the camera parameters

model trained on NuScenes

I can't reproduce the precision by using NuScenes. Can anybody offer me the model trained by Nuscenes. Thanks very much

RuntimeError: cannot perform reduction function argmax on a tensor with no elements because the operation does not have an identity

`Error executing job with overrides: ['+experiments=dd3d_kitti_dla34_overfit', 'EVAL_ONLY=True', 'MODEL.CKPT=/home/cgim/yyh/github/dd3d/model_final.pth', 'TEST.IMS_PER_BATCH=4']
Traceback (most recent call last):
File "./scripts/train.py", line 60, in main
test_results = do_test(cfg, model, is_last=True)
File "./scripts/train.py", line 232, in do_test
per_dataset_results = inference_on_dataset(model, dataloader, evaluator)
File "/home/cgim/anaconda3/envs/yyh/lib/python3.6/site-packages/detectron2/evaluation/evaluator.py", line 158, in inference_on_dataset
outputs = model(inputs)
File "/home/cgim/anaconda3/envs/yyh/lib/python3.6/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/cgim/yyh/github/dd3d/tridet/modeling/dd3d/core.py", line 121, in forward
fcos2d_info
File "/home/cgim/yyh/github/dd3d/tridet/modeling/dd3d/fcos3d.py", line 327, in call
pred_instances[lvl], fcos2d_info[lvl]
File "/home/cgim/yyh/github/dd3d/tridet/modeling/dd3d/fcos3d.py", line 408, in forward_for_single_feature_map
depth_is_distance=self.predict_distance
File "/home/cgim/yyh/github/dd3d/tridet/modeling/dd3d/fcos3d.py", line 48, in predictions_to_boxes3d
quat = allocentric_to_egocentric(quat, proj_ctr, inv_intrinsics)
File "/home/cgim/yyh/github/dd3d/tridet/utils/geometry.py", line 47, in allocentric_to_egocentric
egocentric_quat = matrix_to_quaternion(R_obj_to_global)
File "/home/cgim/anaconda3/envs/yyh/lib/python3.6/site-packages/pytorch3d/transforms/rotation_conversions.py", line 151, in matrix_to_quaternion
F.one_hot(q_abs.argmax(dim=-1), num_classes=4) > 0.5, : # pyre-ignore[16]
RuntimeError: cannot perform reduction function argmax on a tensor with no elements because the operation does not have an identity

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.`

Can you tell me how to solve this problem? When I use this model I get an error like this

OmniML trained weights link not working

For KITTI, OmniML weights link is broken: https://tri-ml-public.s3.amazonaws.com/github/dd3d/experiments/DD3D-OmniML-kitti.pth

What meaning of @ in defaults.yaml?

Can someone tell me the meaning of @ in defaults.yaml？ I can't found its role.
Thanks!

AttributeError: 'DD3D' object has no attribute 'export'

这类问题应该怎么操作呢

Could release other vovnet backbone pre-trained on DDAD15M?

Thanks for your great work! You have released V2-99 backbones pre-trained on DDAD15M using dense depth estimation as the task, and could you release other vovnet backbone pre-trained on DDAD15M?, like VoVNetV2-19, VoVNetV2-39 etc. Thanks a lot!

where to change the train class num 5 to 3

Hi, I find that the dd3d is trained with 5 classes: Car，Pedestrian，Cyclist，Van，Truck , But I just want to train 3 classes (Car，Pedestrian，Cyclist). where to modify the code to train 3 classes ?
I am looking forward to your reply! Thank you very much!

when load pretrained model of VoVNet99, missing key in state_dict

hi, thanks for your work,but when I want to load the pretrained model, the key is not as the same as the model I build.
I extract the model building process into one python file, as below:

# Copyright (c) Youngwan Lee (ETRI) All Rights Reserved.
# Copyright 2021 Toyota Research Institute.  All rights reserved.
from collections import OrderedDict

# import fvcore.nn.weight_init as weight_init
import torch
import torch.nn as nn
import torch.nn.functional as F

from detectron2.layers import FrozenBatchNorm2d, ShapeSpec, get_norm
from detectron2.modeling.backbone import Backbone
from detectron2.modeling.backbone.build import BACKBONE_REGISTRY
from detectron2.modeling.backbone.fpn import FPN, LastLevelMaxPool
import pdb 
    
__all__ = ["VoVNet", "build_vovnet_backbone", "build_vovnet_fpn_backbone"]

_NORM = False

VoVNet19_slim_dw_eSE = {
    'stem': [64, 64, 64],
    'stage_conv_ch': [64, 80, 96, 112],
    'stage_out_ch': [112, 256, 384, 512],
    "layer_per_block": 3,
    "block_per_stage": [1, 1, 1, 1],
    "eSE": True,
    "dw": True
}

VoVNet19_dw_eSE = {
    'stem': [64, 64, 64],
    "stage_conv_ch": [128, 160, 192, 224],
    "stage_out_ch": [256, 512, 768, 1024],
    "layer_per_block": 3,
    "block_per_stage": [1, 1, 1, 1],
    "eSE": True,
    "dw": True
}

VoVNet19_slim_eSE = {
    'stem': [64, 64, 128],
    'stage_conv_ch': [64, 80, 96, 112],
    'stage_out_ch': [112, 256, 384, 512],
    'layer_per_block': 3,
    'block_per_stage': [1, 1, 1, 1],
    'eSE': True,
    "dw": False
}

VoVNet19_eSE = {
    'stem': [64, 64, 128],
    "stage_conv_ch": [128, 160, 192, 224],
    "stage_out_ch": [256, 512, 768, 1024],
    "layer_per_block": 3,
    "block_per_stage": [1, 1, 1, 1],
    "eSE": True,
    "dw": False
}

VoVNet39_eSE = {
    'stem': [64, 64, 128],
    "stage_conv_ch": [128, 160, 192, 224],
    "stage_out_ch": [256, 512, 768, 1024],
    "layer_per_block": 5,
    "block_per_stage": [1, 1, 2, 2],
    "eSE": True,
    "dw": False
}

VoVNet57_eSE = {
    'stem': [64, 64, 128],
    "stage_conv_ch": [128, 160, 192, 224],
    "stage_out_ch": [256, 512, 768, 1024],
    "layer_per_block": 5,
    "block_per_stage": [1, 1, 4, 3],
    "eSE": True,
    "dw": False
}

VoVNet99_eSE = {
    'stem': [64, 64, 128],
    "stage_conv_ch": [128, 160, 192, 224],
    "stage_out_ch": [256, 512, 768, 1024],
    "layer_per_block": 5,
    "block_per_stage": [1, 3, 9, 3],
    "eSE": True,
    "dw": False
}

_STAGE_SPECS = {
    "V-19-slim-dw-eSE": VoVNet19_slim_dw_eSE,
    "V-19-dw-eSE": VoVNet19_dw_eSE,
    "V-19-slim-eSE": VoVNet19_slim_eSE,
    "V-19-eSE": VoVNet19_eSE,
    "V-39-eSE": VoVNet39_eSE,
    "V-57-eSE": VoVNet57_eSE,
    "V-99-eSE": VoVNet99_eSE,
}


def dw_conv3x3(in_channels, out_channels, module_name, postfix, stride=1, kernel_size=3, padding=1):
    """3x3 convolution with padding"""
    return [
        (
            '{}_{}/dw_conv3x3'.format(module_name, postfix),
            nn.Conv2d(
                in_channels,
                out_channels,
                kernel_size=kernel_size,
                stride=stride,
                padding=padding,
                groups=out_channels,
                bias=False
            )
        ),
        (
            '{}_{}/pw_conv1x1'.format(module_name, postfix),
            nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=1, padding=0, groups=1, bias=False)
        ),
        ('{}_{}/pw_norm'.format(module_name, postfix), get_norm(_NORM, out_channels)),
        ('{}_{}/pw_relu'.format(module_name, postfix), nn.ReLU(inplace=True)),
    ]


def conv3x3(in_channels, out_channels, module_name, postfix, stride=1, groups=1, kernel_size=3, padding=1):
    """3x3 convolution with padding"""
    return [
        (
            f"{module_name}_{postfix}/conv",
            nn.Conv2d(
                in_channels,
                out_channels,
                kernel_size=kernel_size,
                stride=stride,
                padding=padding,
                groups=groups,
                bias=False,
            ),
        ),
        (f"{module_name}_{postfix}/norm", get_norm(_NORM, out_channels)),
        (f"{module_name}_{postfix}/relu", nn.ReLU(inplace=True)),
    ]


def conv1x1(in_channels, out_channels, module_name, postfix, stride=1, groups=1, kernel_size=1, padding=0):
    """1x1 convolution with padding"""
    return [
        (
            f"{module_name}_{postfix}/conv",
            nn.Conv2d(
                in_channels,
                out_channels,
                kernel_size=kernel_size,
                stride=stride,
                padding=padding,
                groups=groups,
                bias=False,
            ),
        ),
        (f"{module_name}_{postfix}/norm", get_norm(_NORM, out_channels)),
        (f"{module_name}_{postfix}/relu", nn.ReLU(inplace=True)),
    ]


class Hsigmoid(nn.Module):
    def __init__(self, inplace=True):
        super(Hsigmoid, self).__init__()
        self.inplace = inplace

    def forward(self, x):
        return F.relu6(x + 3.0, inplace=self.inplace) / 6.0


class eSEModule(nn.Module):
    def __init__(self, channel, reduction=4):
        super(eSEModule, self).__init__()
        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.fc = nn.Conv2d(channel, channel, kernel_size=1, padding=0)
        self.hsigmoid = Hsigmoid()

    def forward(self, x):
        input = x
        x = self.avg_pool(x)
        x = self.fc(x)
        x = self.hsigmoid(x)
        return input * x


class _OSA_module(nn.Module):
    def __init__(
        self, in_ch, stage_ch, concat_ch, layer_per_block, module_name, SE=False, identity=False, depthwise=False
    ):

        super(_OSA_module, self).__init__()

        self.identity = identity
        self.depthwise = depthwise
        self.isReduced = False
        self.layers = nn.ModuleList()
        in_channel = in_ch
        if self.depthwise and in_channel != stage_ch:
            self.isReduced = True
            self.conv_reduction = nn.Sequential(
                OrderedDict(conv1x1(in_channel, stage_ch, "{}_reduction".format(module_name), "0"))
            )
        for i in range(layer_per_block):
            if self.depthwise:
                self.layers.append(nn.Sequential(OrderedDict(dw_conv3x3(stage_ch, stage_ch, module_name, i))))
            else:
                self.layers.append(nn.Sequential(OrderedDict(conv3x3(in_channel, stage_ch, module_name, i))))
            in_channel = stage_ch

        # feature aggregation
        in_channel = in_ch + layer_per_block * stage_ch
        self.concat = nn.Sequential(OrderedDict(conv1x1(in_channel, concat_ch, module_name, "concat")))

        self.ese = eSEModule(concat_ch)

    def forward(self, x):

        identity_feat = x

        output = []
        output.append(x)
        if self.depthwise and self.isReduced:
            x = self.conv_reduction(x)
        for layer in self.layers:
            x = layer(x)
            output.append(x)

        x = torch.cat(output, dim=1)
        xt = self.concat(x)

        xt = self.ese(xt)

        if self.identity:
            xt = xt + identity_feat

        return xt


class _OSA_stage(nn.Sequential):
    def __init__(
        self, in_ch, stage_ch, concat_ch, block_per_stage, layer_per_block, stage_num, SE=False, depthwise=False
    ):

        super(_OSA_stage, self).__init__()

        if not stage_num == 2:
            self.add_module("Pooling", nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True))

        if block_per_stage != 1:
            SE = False
        module_name = f"OSA{stage_num}_1"
        self.add_module(
            module_name, _OSA_module(in_ch, stage_ch, concat_ch, layer_per_block, module_name, SE, depthwise=depthwise)
        )
        for i in range(block_per_stage - 1):
            if i != block_per_stage - 2:  # last block
                SE = False
            module_name = f"OSA{stage_num}_{i + 2}"
            self.add_module(
                module_name,
                _OSA_module(
                    concat_ch,
                    stage_ch,
                    concat_ch,
                    layer_per_block,
                    module_name,
                    SE,
                    identity=True,
                    depthwise=depthwise
                ),
            )


class VoVNet(Backbone):
    def __init__(self, cfg, input_ch, out_features=None):
        """
        Args:
            input_ch(int) : the number of input channel
            out_features (list[str]): name of the layers whose outputs should
                be returned in forward. Can be anything in "stem", "stage2" ...
        """
        super(VoVNet, self).__init__()

        global _NORM
        _NORM = cfg.NORM

        stage_specs = _STAGE_SPECS[cfg.NAME]

        stem_ch = stage_specs["stem"]
        config_stage_ch = stage_specs["stage_conv_ch"]
        config_concat_ch = stage_specs["stage_out_ch"]
        block_per_stage = stage_specs["block_per_stage"]
        layer_per_block = stage_specs["layer_per_block"]
        SE = stage_specs["eSE"]
        depthwise = stage_specs["dw"]

        self._out_features = out_features

        # Stem module
        conv_type = dw_conv3x3 if depthwise else conv3x3
        stem = conv3x3(input_ch, stem_ch[0], "stem", "1", 2)
        stem += conv_type(stem_ch[0], stem_ch[1], "stem", "2", 1)
        stem += conv_type(stem_ch[1], stem_ch[2], "stem", "3", 2)
        self.add_module("stem", nn.Sequential((OrderedDict(stem))))
        current_stirde = 4
        self._out_feature_strides = {"stem": current_stirde, "stage2": current_stirde}
        self._out_feature_channels = {"stem": stem_ch[2]}

        stem_out_ch = [stem_ch[2]]
        in_ch_list = stem_out_ch + config_concat_ch[:-1]
        # OSA stages
        self.stage_names = []
        for i in range(4):  # num_stages
            name = "stage%d" % (i + 2)  # stage 2 ... stage 5
            self.stage_names.append(name)
            self.add_module(
                name,
                _OSA_stage(
                    in_ch_list[i],
                    config_stage_ch[i],
                    config_concat_ch[i],
                    block_per_stage[i],
                    layer_per_block,
                    i + 2,
                    SE,
                    depthwise,
                ),
            )

            self._out_feature_channels[name] = config_concat_ch[i]
            if not i == 0:
                self._out_feature_strides[name] = current_stirde = int(current_stirde * 2)

        # initialize weights
        self._initialize_weights()

    def _initialize_weights(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight)

    def _freeze_backbone(self, freeze_at):
        if freeze_at < 0:
            return

        for stage_index in range(freeze_at):
            if stage_index == 0:
                m = self.stem  # stage 0 is the stem
            else:
                m = getattr(self, "stage" + str(stage_index + 1))
            for p in m.parameters():
                p.requires_grad = False
                FrozenBatchNorm2d.convert_frozen_batchnorm(self)

    def forward(self, x):
        outputs = {}
        x = self.stem(x)
        if "stem" in self._out_features:
            outputs["stem"] = x
        for name in self.stage_names:
            x = getattr(self, name)(x)
            if name in self._out_features:
                outputs[name] = x

        return outputs

    def output_shape(self):
        return {
            name: ShapeSpec(channels=self._out_feature_channels[name], stride=self._out_feature_strides[name])
            for name in self._out_features
        }


@BACKBONE_REGISTRY.register()
def build_vovnet_backbone(cfg, input_shape):
    """
    Create a VoVNet instance from config.
    Returns:
        VoVNet: a :class:`VoVNet` instance.
    """
    out_features = cfg.OUT_FEATURES
    return VoVNet(cfg, input_shape.channels, out_features=out_features)


@BACKBONE_REGISTRY.register()
def build_vovnet_fpn_backbone(cfg, input_shape: ShapeSpec):
    """
    Args:
        cfg: a detectron2 CfgNode
    Returns:
        backbone (Backbone): backbone module, must be a subclass of :class:`Backbone`.
    """
    bottom_up = build_vovnet_backbone(cfg, input_shape)
    in_features = cfg.MODEL.FPN.IN_FEATURES
    out_channels = cfg.MODEL.FPN.OUT_CHANNELS
    backbone = FPN(
        bottom_up=bottom_up,
        in_features=in_features,
        out_channels=out_channels,
        norm=cfg.FE.FPN.NORM,
        top_block=LastLevelMaxPool(),
        fuse_type=cfg.FE.FPN.FUSE_TYPE,
    )
    return backbone


# class LastLevelP6(nn.Module):
#     """
#     This module is used in FCOS to generate extra layers
#     """
#     def __init__(self, in_channels, out_channels, in_features="res5"):
#         super().__init__()
#         self.num_levels = 1
#         self.in_feature = in_features
#         self.p6 = nn.Conv2d(in_channels, out_channels, 3, 2, 1)
#         for module in [self.p6]:
#             weight_init.c2_xavier_fill(module)

#     def forward(self, x):
#         p6 = self.p6(x)
#         return [p6]


@BACKBONE_REGISTRY.register()
def build_fcos_vovnet_fpn_backbone_p6(cfg, input_shape: ShapeSpec):
    """
    Args:
        cfg: a detectron2 CfgNode
    Returns:
        backbone (Backbone): backbone module, must be a subclass of :class:`Backbone`.
    """
    bottom_up = build_vovnet_backbone(cfg.FE.BACKBONE, input_shape)
    in_features = cfg.FE.FPN.IN_FEATURES
    out_channels = cfg.FE.FPN.OUT_CHANNELS
    in_channels_top = out_channels
    top_block = LastLevelP6(in_channels_top, out_channels, "p5")
    backbone = FPN(
        bottom_up=bottom_up,
        in_features=in_features,
        out_channels=out_channels,
        norm=cfg.FE.FPN.NORM,
        # top_block=LastLevelMaxPool(),
        top_block=top_block,
        fuse_type=cfg.FE.FPN.FUSE_TYPE,
    )

    backbone._size_divisibility *= 2

    return backbone

import argparse
import yaml
import os
from yacs.config import CfgNode as CN
_C = CN()
_C.defaults=['d2_fpn@FPN']

_C.BUILDER='build_fcos_vovnet_fpn_backbone_p6'

_C.BACKBONE=CN()

_C.BACKBONE.NAME= 'V-99-eSE'
_C.BACKBONE.OUT_FEATURES=['stage2', 'stage3', 'stage4', 'stage5']
_C.BACKBONE.NORM='BN'

_C.NAME= 'V-99-eSE'
_C.OUT_FEATURES = ['stage2', 'stage3', 'stage4', 'stage5']
_C.NORM='BN'

_C.MODEL=CN()
_C.MODEL.BACKBONE=CN()
_C.MODEL.BACKBONE.NORM = 'FrozenBN'
_C.MODEL.FPN=CN()
_C.MODEL.FPN.NORM = 'FrozenBN'
_C.MODEL.FPN.IN_FEATURES= ['stage2', 'stage3', 'stage4', 'stage5']#["level3", "level4", "level5"]
_C.MODEL.FPN.OUT_CHANNELS=256

_C.FE=CN()
_C.FE.BACKBONE=CN()
_C.FE.BACKBONE.NORM='FrozenBN'
_C.FE.FPN=CN()
_C.FE.FPN.NORM='FrozenBN'
_C.FE.FPN.FUSE_TYPE='sum'#sum/avg
_C.FE.OUT_FEATURES= _C.OUT_FEATURES


def parse_option():
    parser = argparse.ArgumentParser('VoVNet backbone config', add_help=False)
    parser.add_argument('--cfg', type=str, required=False, default = './vovnet_cfg/v2_99_fpn.yaml',metavar="FILE", help='path to config file', )
    parser.add_argument(
        "--opts",
        help="Modify config options by adding 'KEY VALUE' pairs. ",
        default=None,
        nargs='+',
    )
    args, unparsed = parser.parse_known_args()

    config = get_config(args)
    return args, config

def _update_config_from_file(config, cfg_file):
    config.defrost()
    with open(cfg_file, 'r') as f:
        yaml_cfg = yaml.load(f, Loader=yaml.FullLoader)

    for cfg in yaml_cfg.setdefault('BASE', ['']):
        if cfg:
            _update_config_from_file(
                config, os.path.join(os.path.dirname(cfg_file), cfg)
            )
    print('=> merge config from {}'.format(cfg_file))
    config.merge_from_file(cfg_file)
    config.freeze()

def update_config(config, args):
    _update_config_from_file(config, args.cfg)
    config.defrost()
    if args.opts:
        config.merge_from_list(args.opts)
    config.freeze()

def get_config(args):
    """Get a yacs CfgNode object with default values."""
    # Return a clone so that the defaults will not be altered
    # This is for the "local variable" use pattern
    config = _C.clone()
    update_config(config, args)

    return config
if __name__ == '__main__':
    args, config = parse_option()
    
    model = build_vovnet_fpn_backbone(config,input_shape = ShapeSpec(channels=3, height=448, width=800, stride=None))
    # pretrained_model = torch.load('/data/CenterFusion0/models/vovnet99_dd3d15m.pth',map_location=torch.device('cpu'))
    # model.load_state_dict(pretrained_model)
    x = torch.randn(1,3,448,800)
    y = model.forward(x)
    pdb.set_trace()
    print(y)

The error info is shown as below:

/home/gongzheng/.local/lib/python3.6/site-packages/redis/utils.py:12: CryptographyDeprecationWarning: Python 3.6 is no longer supported by the Python core team. Therefo
re, support for it is deprecated in cryptography and will be removed in a future release.                                                                               
  import cryptography  # noqa                                                                                                                                           
=> merge config from ./vovnet_cfg/v2_99_fpn.yaml                                                                                                                        
Traceback (most recent call last):                                                                                                                                      
  File "vovnet.py", line 538, in <module>                                                                                                                               
    model.load_state_dict(pretrained_model)                                                                                                                             
  File "/home/gongzheng/anaconda3/envs/cf/lib/python3.6/site-packages/torch/nn/modules/module.py", line 1483, in load_state_dict                                        
    self.__class__.__name__, "\n\t".join(error_msgs)))                                                                                                                  
RuntimeError: Error(s) in loading state_dict for FPN:                                                                                                                   
        Missing key(s) in state_dict: "fpn_lateral2.weight", "fpn_lateral2.norm.weight", "fpn_lateral2.norm.bias", "fpn_output2.weight", "fpn_output2.norm.weight", "fpn
_output2.norm.bias", "fpn_lateral3.weight", "fpn_lateral3.norm.weight", "fpn_lateral3.norm.bias", "fpn_output3.weight", "fpn_output3.norm.weight", "fpn_output3.norm.bia
s", "fpn_lateral4.weight", "fpn_lateral4.norm.weight", "fpn_lateral4.norm.bias", "fpn_output4.weight", "fpn_output4.norm.weight", "fpn_output4.norm.bias", "fpn_lateral5
.weight", "fpn_lateral5.norm.weight", "fpn_lateral5.norm.bias", "fpn_output5.weight", "fpn_output5.norm.weight", "fpn_output5.norm.bias", "bottom_up.stem.stem_1/conv.we
ight", "bottom_up.stem.stem_1/norm.weight", "bottom_up.stem.stem_1/norm.bias", "bottom_up.stem.stem_1/norm.running_mean", "bottom_up.stem.stem_1/norm.running_var", "bot
tom_up.stem.stem_2/conv.weight", "bottom_up.stem.stem_2/norm.weight", "bottom_up.stem.stem_2/norm.bias", "bottom_up.stem.stem_2/norm.running_mean", "bottom_up.stem.stem
_2/norm.running_var", "bottom_up.stem.stem_3/conv.weight", "bottom_up.stem.stem_3/norm.weight", "bottom_up.stem.stem_3/norm.bias", "bottom_up.stem.stem_3/norm.running_m
ean", "bottom_up.stem.stem_3/norm.running_var", "bottom_up.stage2.OSA2_1.layers.0.OSA2_1_0/conv.weight", "bottom_up.stage2.OSA2_1.layers.0.OSA2_1_0/norm.weight", "botto
m_up.stage2.OSA2_1.layers.0.OSA2_1_0/norm.bias", "bottom_up.stage2.OSA2_1.layers.0.OSA2_1_0/norm.running_mean", "bottom_up.stage2.OSA2_1.layers.0.OSA2_1_0/norm.running_
var", "bottom_up.stage2.OSA2_1.layers.1.OSA2_1_1/conv.weight", "bottom_up.stage2.OSA2_1.layers.1.OSA2_1_1/norm.weight", "bottom_up.stage2.OSA2_1.layers.1.OSA2_1_1/norm.
bias", "bottom_up.stage2.OSA2_1.layers.1.OSA2_1_1/norm.running_mean", "bottom_up.stage2.OSA2_1.layers.1.OSA2_1_1/norm.running_var", "bottom_up.stage2.OSA2_1.layers.2.OS
A2_1_2/conv.weight", "bottom_up.stage2.OSA2_1.layers.2.OSA2_1_2/norm.weight", "bottom_up.stage2.OSA2_1.layers.2.OSA2_1_2/norm.bias", "bottom_up.stage2.OSA2_1.layers.2.O
SA2_1_2/norm.running_mean", "bottom_up.stage2.OSA2_1.layers.2.OSA2_1_2/norm.running_var", "bottom_up.stage2.OSA2_1.layers.3.OSA2_1_3/conv.weight", "bottom_up.stage2.OSA
2_1.layers.3.OSA2_1_3/norm.weight", "bottom_up.stage2.OSA2_1.layers.3.OSA2_1_3/norm.bias", "bottom_up.stage2.OSA2_1.layers.3.OSA2_1_3/norm.running_mean", "bottom_up.sta
ge2.OSA2_1.layers.3.OSA2_1_3/norm.running_var", "bottom_up.stage2.OSA2_1.layers.4.OSA2_1_4/conv.weight", "bottom_up.stage2.OSA2_1.layers.4.OSA2_1_4/norm.weight", "botto
m_up.stage2.OSA2_1.layers.4.OSA2_1_4/norm.bias", "bottom_up.stage2.OSA2_1.layers.4.OSA2_1_4/norm.running_mean", "bottom_up.stage2.OSA2_1.layers.4.OSA2_1_4/norm.running_
var", "bottom_up.stage2.OSA2_1.concat.OSA2_1_concat/conv.weight", "bottom_up.stage2.OSA2_1.concat.OSA2_1_concat/norm.weight", "bottom_up.stage2.OSA2_1.concat.OSA2_1_con
cat/norm.bias", "bottom_up.stage2.OSA2_1.concat.OSA2_1_concat/norm.running_mean", "bottom_up.stage2.OSA2_1.concat.OSA2_1_concat/norm.running_var", "bottom_up.stage2.OSA
2_1.ese.fc.weight", "bottom_up.stage2.OSA2_1.ese.fc.bias", "bottom_up.stage3.OSA3_1.layers.0.OSA3_1_0/conv.weight", "bottom_up.stage3.OSA3_1.layers.0.OSA3_1_0/norm.weig
ht", "bottom_up.stage3.OSA3_1.layers.0.OSA3_1_0/norm.bias", "bottom_up.stage3.OSA3_1.layers.0.OSA3_1_0/norm.running_mean", "bottom_up.stage3.OSA3_1.layers.0.OSA3_1_0/no
rm.running_var", "bottom_up.stage3.OSA3_1.layers.1.OSA3_1_1/conv.weight", "bottom_up.stage3.OSA3_1.layers.1.OSA3_1_1/norm.weight", "bottom_up.stage3.OSA3_1.layers.1.OSA
3_1_1/norm.bias", "bottom_up.stage3.OSA3_1.layers.1.OSA3_1_1/norm.running_mean", "bottom_up.stage3.OSA3_1.layers.1.OSA3_1_1/norm.running_var", "bottom_up.stage3.OSA3_1.
layers.2.OSA3_1_2/conv.weight", "bottom_up.stage3.OSA3_1.layers.2.OSA3_1_2/norm.weight", "bottom_up.stage3.OSA3_1.layers.2.OSA3_1_2/norm.bias", "bottom_up.stage3.OSA3_1
.layers.2.OSA3_1_2/norm.running_mean", "bottom_up.stage3.OSA3_1.layers.2.OSA3_1_2/norm.running_var", "bottom_up.stage3.OSA3_1.layers.3.OSA3_1_3/conv.weight", "bottom_up
.stage3.OSA3_1.layers.3.OSA3_1_3/norm.weight", "bottom_up.stage3.OSA3_1.layers.3.OSA3_1_3/norm.bias", "bottom_up.stage3.OSA3_1.layers.3.OSA3_1_3/norm.running_mean", "bo
ttom_up.stage3.OSA3_1.layers.3.OSA3_1_3/norm.running_var", "bottom_up.stage3.OSA3_1.layers.4.OSA3_1_4/conv.weight", "bottom_up.stage3.OSA3_1.layers.4.OSA3_1_4/norm.weig
ht", "bottom_up.stage3.OSA3_1.layers.4.OSA3_1_4/norm.bias", "bottom_up.stage3.OSA3_1.layers.4.OSA3_1_4/norm.running_mean", "bottom_up.stage3.OSA3_1.layers.4.OSA3_1_4/no
rm.running_var", "bottom_up.stage3.OSA3_1.concat.OSA3_1_concat/conv.weight", "bottom_up.stage3.OSA3_1.concat.OSA3_1_concat/norm.weight", "bottom_up.stage3.OSA3_1.concat
.OSA3_1_concat/norm.bias", "bottom_up.stage3.OSA3_1.concat.OSA3_1_concat/norm.running_mean", "bottom_up.stage3.OSA3_1.concat.OSA3_1_concat/norm.running_var", "bottom_up
.stage3.OSA3_1.ese.fc.weight", "bottom_up.stage3.OSA3_1.ese.fc.bias", "bottom_up.stage3.OSA3_2.layers.0.OSA3_2_0/conv.weight", "bottom_up.stage3.OSA3_2.layers.0.OSA3_2_
0/norm.weight", "bottom_up.stage3.OSA3_2.layers.0.OSA3_2_0/norm.bias", "bottom_up.stage3.OSA3_2.layers.0.OSA3_2_0/norm.running_mean", "bottom_up.stage3.OSA3_2.layers.0.
OSA3_2_0/norm.running_var", "bottom_up.stage3.OSA3_2.layers.1.OSA3_2_1/conv.weight", "bottom_up.stage3.OSA3_2.layers.1.OSA3_2_1/norm.weight", "bottom_up.stage3.OSA3_2.l
ayers.1.OSA3_2_1/norm.bias", "bottom_up.stage3.OSA3_2.layers.1.OSA3_2_1/norm.running_mean", "bottom_up.stage3.OSA3_2.layers.1.OSA3_2_1/norm.running_var", "bottom_up.sta
ge3.OSA3_2.layers.2.OSA3_2_2/conv.weight", "bottom_up.stage3.OSA3_2.layers.2.OSA3_2_2/norm.weight", "bottom_up.stage3.OSA3_2.layers.2.OSA3_2_2/norm.bias", "bottom_up.st
age3.OSA3_2.layers.2.OSA3_2_2/norm.running_mean", "bottom_up.stage3.OSA3_2.layers.2.OSA3_2_2/norm.running_var", "bottom_up.stage3.OSA3_2.layers.3.OSA3_2_3/conv.weight",
 "bottom_up.stage3.OSA3_2.layers.3.OSA3_2_3/norm.weight", "bottom_up.stage3.OSA3_2.layers.3.OSA3_2_3/norm.bias", "bottom_up.stage3.OSA3_2.layers.3.OSA3_2_3/norm.running_mean", "bottom_up.stage3.OSA3_2.layers.3.OSA3_2_3/norm.running_var", "bottom_up.stage3.OSA3_2.layers.4.OSA3_2_4/conv.weight", "bottom_up.stage3.OSA3_2.layers.4.OSA3_2_
4/norm.weight", "bottom_up.stage3.OSA3_2.layers.4.OSA3_2_4/norm.bias", "bottom_up.stage3.OSA3_2.layers.4.OSA3_2_4/norm.running_mean", "bottom_up.stage3.OSA3_2.layers.4.
OSA3_2_4/norm.running_var", "bottom_up.stage3.OSA3_2.concat.OSA3_2_concat/conv.weight", "bottom_up.stage3.OSA3_2.concat.OSA3_2_concat/norm.weight", "bottom_up.stage3.OS
A3_2.concat.OSA3_2_concat/norm.bias", "bottom_up.stage3.OSA3_2.concat.OSA3_2_concat/norm.running_mean", "bottom_up.stage3.OSA3_2.concat.OSA3_2_concat/norm.running_var",
 "bottom_up.stage3.OSA3_2.ese.fc.weight", "bottom_up.stage3.OSA3_2.ese.fc.bias", "bottom_up.stage3.OSA3_3.layers.0.OSA3_3_0/conv.weight", "bottom_up.stage3.OSA3_3.layer
s.0.OSA3_3_0/norm.weight", "bottom_up.stage3.OSA3_3.layers.0.OSA3_3_0/norm.bias", "bottom_up.stage3.OSA3_3.layers.0.OSA3_3_0/norm.running_mean", "bottom_up.stage3.OSA3_
3.layers.0.OSA3_3_0/norm.running_var", "bottom_up.stage3.OSA3_3.layers.1.OSA3_3_1/conv.weight", "bottom_up.stage3.OSA3_3.layers.1.OSA3_3_1/norm.weight", "bottom_up.stag
e3.OSA3_3.layers.1.OSA3_3_1/norm.bias", "bottom_up.stage3.OSA3_3.layers.1.OSA3_3_1/norm.running_mean", "bottom_up.stage3.OSA3_3.layers.1.OSA3_3_1/norm.running_var", "bo
ttom_up.stage3.OSA3_3.layers.2.OSA3_3_2/conv.weight", "bottom_up.stage3.OSA3_3.layers.2.OSA3_3_2/norm.weight", "bottom_up.stage3.OSA3_3.layers.2.OSA3_3_2/norm.bias", "b
ottom_up.stage3.OSA3_3.layers.2.OSA3_3_2/norm.running_mean", "bottom_up.stage3.OSA3_3.layers.2.OSA3_3_2/norm.running_var", "bottom_up.stage3.OSA3_3.layers.3.OSA3_3_3/co
nv.weight", "bottom_up.stage3.OSA3_3.layers.3.OSA3_3_3/norm.weight", "bottom_up.stage3.OSA3_3.layers.3.OSA3_3_3/norm.bias", "bottom_up.stage3.OSA3_3.layers.3.OSA3_3_3/n
orm.running_mean", "bottom_up.stage3.OSA3_3.layers.3.OSA3_3_3/norm.running_var", "bottom_up.stage3.OSA3_3.layers.4.OSA3_3_4/conv.weight", "bottom_up.stage3.OSA3_3.layer
s.4.OSA3_3_4/norm.weight", "bottom_up.stage3.OSA3_3.layers.4.OSA3_3_4/norm.bias", "bottom_up.stage3.OSA3_3.layers.4.OSA3_3_4/norm.running_mean", "bottom_up.stage3.OSA3_
3.layers.4.OSA3_3_4/norm.running_var", "bottom_up.stage3.OSA3_3.concat.OSA3_3_concat/conv.weight", "bottom_up.stage3.OSA3_3.concat.OSA3_3_concat/norm.weight", "bottom_u
p.stage3.OSA3_3.concat.OSA3_3_concat/norm.bias", "bottom_up.stage3.OSA3_3.concat.OSA3_3_concat/norm.running_mean", "bottom_up.stage3.OSA3_3.concat.OSA3_3_concat/norm.ru
nning_var", "bottom_up.stage3.OSA3_3.ese.fc.weight", "bottom_up.stage3.OSA3_3.ese.fc.bias", "bottom_up.stage4.OSA4_1.layers.0.OSA4_1_0/conv.weight", "bottom_up.stage4.O
SA4_1.layers.0.OSA4_1_0/norm.weight", "bottom_up.stage4.OSA4_1.layers.0.OSA4_1_0/norm.bias", "bottom_up.stage4.OSA4_1.layers.0.OSA4_1_0/norm.running_mean", "bottom_up.s
tage4.OSA4_1.layers.0.OSA4_1_0/norm.running_var", "bottom_up.stage4.OSA4_1.layers.1.OSA4_1_1/conv.weight", "bottom_up.stage4.OSA4_1.layers.1.OSA4_1_1/norm.weight", "bot
tom_up.stage4.OSA4_1.layers.1.OSA4_1_1/norm.bias", "bottom_up.stage4.OSA4_1.layers.1.OSA4_1_1/norm.running_mean", "bottom_up.stage4.OSA4_1.layers.1.OSA4_1_1/norm.runnin
g_var", "bottom_up.stage4.OSA4_1.layers.2.OSA4_1_2/conv.weight", "bottom_up.stage4.OSA4_1.layers.2.OSA4_1_2/norm.weight", "bottom_up.stage4.OSA4_1.layers.2.OSA4_1_2/nor
m.bias", "bottom_up.stage4.OSA4_1.layers.2.OSA4_1_2/norm.running_mean", "bottom_up.stage4.OSA4_1.layers.2.OSA4_1_2/norm.running_var", "bottom_up.stage4.OSA4_1.layers.3.
OSA4_1_3/conv.weight", "bottom_up.stage4.OSA4_1.layers.3.OSA4_1_3/norm.weight", "bottom_up.stage4.OSA4_1.layers.3.OSA4_1_3/norm.bias", "bottom_up.stage4.OSA4_1.layers.3
.OSA4_1_3/norm.running_mean", "bottom_up.stage4.OSA4_1.layers.3.OSA4_1_3/norm.running_var", "bottom_up.stage4.OSA4_1.layers.4.OSA4_1_4/conv.weight", "bottom_up.stage4.O
SA4_1.layers.4.OSA4_1_4/norm.weight", "bottom_up.stage4.OSA4_1.layers.4.OSA4_1_4/norm.bias", "bottom_up.stage4.OSA4_1.layers.4.OSA4_1_4/norm.running_mean", "bottom_up.s
tage4.OSA4_1.layers.4.OSA4_1_4/norm.running_var", "bottom_up.stage4.OSA4_1.concat.OSA4_1_concat/conv.weight", "bottom_up.stage4.OSA4_1.concat.OSA4_1_concat/norm.weight"
, "bottom_up.stage4.OSA4_1.concat.OSA4_1_concat/norm.bias", "bottom_up.stage4.OSA4_1.concat.OSA4_1_concat/norm.running_mean", "bottom_up.stage4.OSA4_1.concat.OSA4_1_con
cat/norm.running_var", "bottom_up.stage4.OSA4_1.ese.fc.weight", "bottom_up.stage4.OSA4_1.ese.fc.bias", "bottom_up.stage4.OSA4_2.layers.0.OSA4_2_0/conv.weight", "bottom_
up.stage4.OSA4_2.layers.0.OSA4_2_0/norm.weight", "bottom_up.stage4.OSA4_2.layers.0.OSA4_2_0/norm.bias", "bottom_up.stage4.OSA4_2.layers.0.OSA4_2_0/norm.running_mean", "
bottom_up.stage4.OSA4_2.layers.0.OSA4_2_0/norm.running_var", "bottom_up.stage4.OSA4_2.layers.1.OSA4_2_1/conv.weight", "bottom_up.stage4.OSA4_2.layers.1.OSA4_2_1/norm.we
ight", "bottom_up.stage4.OSA4_2.layers.1.OSA4_2_1/norm.bias", "bottom_up.stage4.OSA4_2.layers.1.OSA4_2_1/norm.running_mean", "bottom_up.stage4.OSA4_2.layers.1.OSA4_2_1/
norm.running_var", "bottom_up.stage4.OSA4_2.layers.2.OSA4_2_2/conv.weight", "bottom_up.stage4.OSA4_2.layers.2.OSA4_2_2/norm.weight", "bottom_up.stage4.OSA4_2.layers.2.O
SA4_2_2/norm.bias", "bottom_up.stage4.OSA4_2.layers.2.OSA4_2_2/norm.running_mean", "bottom_up.stage4.OSA4_2.layers.2.OSA4_2_2/norm.running_var", "bottom_up.stage4.OSA4_
2.layers.3.OSA4_2_3/conv.weight", "bottom_up.stage4.OSA4_2.layers.3.OSA4_2_3/norm.weight", "bottom_up.stage4.OSA4_2.layers.3.OSA4_2_3/norm.bias", "bottom_up.stage4.OSA4
_2.layers.3.OSA4_2_3/norm.running_mean", "bottom_up.stage4.OSA4_2.layers.3.OSA4_2_3/norm.running_var", "bottom_up.stage4.OSA4_2.layers.4.OSA4_2_4/conv.weight", "bottom_
up.stage4.OSA4_2.layers.4.OSA4_2_4/norm.weight", "bottom_up.stage4.OSA4_2.layers.4.OSA4_2_4/norm.bias", "bottom_up.stage4.OSA4_2.layers.4.OSA4_2_4/norm.running_mean", "
bottom_up.stage4.OSA4_2.layers.4.OSA4_2_4/norm.running_var", "bottom_up.stage4.OSA4_2.concat.OSA4_2_concat/conv.weight", "bottom_up.stage4.OSA4_2.concat.OSA4_2_concat/n
orm.weight", "bottom_up.stage4.OSA4_2.concat.OSA4_2_concat/norm.bias", "bottom_up.stage4.OSA4_2.concat.OSA4_2_concat/norm.running_mean", "bottom_up.stage4.OSA4_2.concat
.OSA4_2_concat/norm.running_var", "bottom_up.stage4.OSA4_2.ese.fc.weight", "bottom_up.stage4.OSA4_2.ese.fc.bias", "bottom_up.stage4.OSA4_3.layers.0.OSA4_3_0/conv.weight
", "bottom_up.stage4.OSA4_3.layers.0.OSA4_3_0/norm.weight", "bottom_up.stage4.OSA4_3.layers.0.OSA4_3_0/norm.bias", "bottom_up.stage4.OSA4_3.layers.0.OSA4_3_0/norm.runni
ng_mean", "bottom_up.stage4.OSA4_3.layers.0.OSA4_3_0/norm.running_var", "bottom_up.stage4.OSA4_3.layers.1.OSA4_3_1/conv.weight", "bottom_up.stage4.OSA4_3.layers.1.OSA4_
3_1/norm.weight", "bottom_up.stage4.OSA4_3.layers.1.OSA4_3_1/norm.bias", "bottom_up.stage4.OSA4_3.layers.1.OSA4_3_1/norm.running_mean", "bottom_up.stage4.OSA4_3.layers.
1.OSA4_3_1/norm.running_var", "bottom_up.stage4.OSA4_3.layers.2.OSA4_3_2/conv.weight", "bottom_up.stage4.OSA4_3.layers.2.OSA4_3_2/norm.weight", "bottom_up.stage4.OSA4_3
.layers.2.OSA4_3_2/norm.bias", "bottom_up.stage4.OSA4_3.layers.2.OSA4_3_2/norm.running_mean", "bottom_up.stage4.OSA4_3.layers.2.OSA4_3_2/norm.running_var", "bottom_up.s
tage4.OSA4_3.layers.3.OSA4_3_3/conv.weight", "bottom_up.stage4.OSA4_3.layers.3.OSA4_3_3/norm.weight", "bottom_up.stage4.OSA4_3.layers.3.OSA4_3_3/norm.bias", "bottom_up.
stage4.OSA4_3.layers.3.OSA4_3_3/norm.running_mean", "bottom_up.stage4.OSA4_3.layers.3.OSA4_3_3/norm.running_var", "bottom_up.stage4.OSA4_3.layers.4.OSA4_3_4/conv.weight
", "bottom_up.stage4.OSA4_3.layers.4.OSA4_3_4/norm.weight", "bottom_up.stage4.OSA4_3.layers.4.OSA4_3_4/norm.bias", "bottom_up.stage4.OSA4_3.layers.4.OSA4_3_4/norm.runni
ng_mean", "bottom_up.stage4.OSA4_3.layers.4.OSA4_3_4/norm.running_var", "bottom_up.stage4.OSA4_3.concat.OSA4_3_concat/conv.weight", "bottom_up.stage4.OSA4_3.concat.OSA4
_3_concat/norm.weight", "bottom_up.stage4.OSA4_3.concat.OSA4_3_concat/norm.bias", "bottom_up.stage4.OSA4_3.concat.OSA4_3_concat/norm.running_mean", "bottom_up.stage4.OS
A4_3.concat.OSA4_3_concat/norm.running_var", "bottom_up.stage4.OSA4_3.ese.fc.weight", "bottom_up.stage4.OSA4_3.ese.fc.bias", "bottom_up.stage4.OSA4_4.layers.0.OSA4_4_0/
conv.weight", "bottom_up.stage4.OSA4_4.layers.0.OSA4_4_0/norm.weight", "bottom_up.stage4.OSA4_4.layers.0.OSA4_4_0/norm.bias", "bottom_up.stage4.OSA4_4.layers.0.OSA4_4_0
/norm.running_mean", "bottom_up.stage4.OSA4_4.layers.0.OSA4_4_0/norm.running_var", "bottom_up.stage4.OSA4_4.layers.1.OSA4_4_1/conv.weight", "bottom_up.stage4.OSA4_4.layers.1.OSA4_4_1/norm.weight", "bottom_up.stage4.OSA4_4.layers.1.OSA4_4_1/norm.bias", "bottom_up.stage4.OSA4_4.layers.1.OSA4_4_1/norm.running_mean", "bottom_up.stage4.OSA
4_4.layers.1.OSA4_4_1/norm.running_var", "bottom_up.stage4.OSA4_4.layers.2.OSA4_4_2/conv.weight", "bottom_up.stage4.OSA4_4.layers.2.OSA4_4_2/norm.weight", "bottom_up.st
age4.OSA4_4.layers.2.OSA4_4_2/norm.bias", "bottom_up.stage4.OSA4_4.layers.2.OSA4_4_2/norm.running_mean", "bottom_up.stage4.OSA4_4.layers.2.OSA4_4_2/norm.running_var", "
bottom_up.stage4.OSA4_4.layers.3.OSA4_4_3/conv.weight", "bottom_up.stage4.OSA4_4.layers.3.OSA4_4_3/norm.weight", "bottom_up.stage4.OSA4_4.layers.3.OSA4_4_3/norm.bias", 
"bottom_up.stage4.OSA4_4.layers.3.OSA4_4_3/norm.running_mean", "bottom_up.stage4.OSA4_4.layers.3.OSA4_4_3/norm.running_var", "bottom_up.stage4.OSA4_4.layers.4.OSA4_4_4/
conv.weight", "bottom_up.stage4.OSA4_4.layers.4.OSA4_4_4/norm.weight", "bottom_up.stage4.OSA4_4.layers.4.OSA4_4_4/norm.bias", "bottom_up.stage4.OSA4_4.layers.4.OSA4_4_4
/norm.running_mean", "bottom_up.stage4.OSA4_4.layers.4.OSA4_4_4/norm.running_var", "bottom_up.stage4.OSA4_4.concat.OSA4_4_concat/conv.weight", "bottom_up.stage4.OSA4_4.
concat.OSA4_4_concat/norm.weight", "bottom_up.stage4.OSA4_4.concat.OSA4_4_concat/norm.bias", "bottom_up.stage4.OSA4_4.concat.OSA4_4_concat/norm.running_mean", "bottom_u
p.stage4.OSA4_4.concat.OSA4_4_concat/norm.running_var", "bottom_up.stage4.OSA4_4.ese.fc.weight", "bottom_up.stage4.OSA4_4.ese.fc.bias", "bottom_up.stage4.OSA4_5.layers.
0.OSA4_5_0/conv.weight", "bottom_up.stage4.OSA4_5.layers.0.OSA4_5_0/norm.weight", "bottom_up.stage4.OSA4_5.layers.0.OSA4_5_0/norm.bias", "bottom_up.stage4.OSA4_5.layers
.0.OSA4_5_0/norm.running_mean", "bottom_up.stage4.OSA4_5.layers.0.OSA4_5_0/norm.running_var", "bottom_up.stage4.OSA4_5.layers.1.OSA4_5_1/conv.weight", "bottom_up.stage4
.OSA4_5.layers.1.OSA4_5_1/norm.weight", "bottom_up.stage4.OSA4_5.layers.1.OSA4_5_1/norm.bias", "bottom_up.stage4.OSA4_5.layers.1.OSA4_5_1/norm.running_mean", "bottom_up
.stage4.OSA4_5.layers.1.OSA4_5_1/norm.running_var", "bottom_up.stage4.OSA4_5.layers.2.OSA4_5_2/conv.weight", "bottom_up.stage4.OSA4_5.layers.2.OSA4_5_2/norm.weight", "b
ottom_up.stage4.OSA4_5.layers.2.OSA4_5_2/norm.bias", "bottom_up.stage4.OSA4_5.layers.2.OSA4_5_2/norm.running_mean", "bottom_up.stage4.OSA4_5.layers.2.OSA4_5_2/norm.runn
ing_var", "bottom_up.stage4.OSA4_5.layers.3.OSA4_5_3/conv.weight", "bottom_up.stage4.OSA4_5.layers.3.OSA4_5_3/norm.weight", "bottom_up.stage4.OSA4_5.layers.3.OSA4_5_3/n
orm.bias", "bottom_up.stage4.OSA4_5.layers.3.OSA4_5_3/norm.running_mean", "bottom_up.stage4.OSA4_5.layers.3.OSA4_5_3/norm.running_var", "bottom_up.stage4.OSA4_5.layers.
4.OSA4_5_4/conv.weight", "bottom_up.stage4.OSA4_5.layers.4.OSA4_5_4/norm.weight", "bottom_up.stage4.OSA4_5.layers.4.OSA4_5_4/norm.bias", "bottom_up.stage4.OSA4_5.layers
.4.OSA4_5_4/norm.running_mean", "bottom_up.stage4.OSA4_5.layers.4.OSA4_5_4/norm.running_var", "bottom_up.stage4.OSA4_5.concat.OSA4_5_concat/conv.weight", "bottom_up.sta
ge4.OSA4_5.concat.OSA4_5_concat/norm.weight", "bottom_up.stage4.OSA4_5.concat.OSA4_5_concat/norm.bias", "bottom_up.stage4.OSA4_5.concat.OSA4_5_concat/norm.running_mean"
, "bottom_up.stage4.OSA4_5.concat.OSA4_5_concat/norm.running_var", "bottom_up.stage4.OSA4_5.ese.fc.weight", "bottom_up.stage4.OSA4_5.ese.fc.bias", "bottom_up.stage4.OSA
4_6.layers.0.OSA4_6_0/conv.weight", "bottom_up.stage4.OSA4_6.layers.0.OSA4_6_0/norm.weight", "bottom_up.stage4.OSA4_6.layers.0.OSA4_6_0/norm.bias", "bottom_up.stage4.OS
A4_6.layers.0.OSA4_6_0/norm.running_mean", "bottom_up.stage4.OSA4_6.layers.0.OSA4_6_0/norm.running_var", "bottom_up.stage4.OSA4_6.layers.1.OSA4_6_1/conv.weight", "botto
m_up.stage4.OSA4_6.layers.1.OSA4_6_1/norm.weight", "bottom_up.stage4.OSA4_6.layers.1.OSA4_6_1/norm.bias", "bottom_up.stage4.OSA4_6.layers.1.OSA4_6_1/norm.running_mean",
 "bottom_up.stage4.OSA4_6.layers.1.OSA4_6_1/norm.running_var", "bottom_up.stage4.OSA4_6.layers.2.OSA4_6_2/conv.weight", "bottom_up.stage4.OSA4_6.layers.2.OSA4_6_2/norm.
weight", "bottom_up.stage4.OSA4_6.layers.2.OSA4_6_2/norm.bias", "bottom_up.stage4.OSA4_6.layers.2.OSA4_6_2/norm.running_mean", "bottom_up.stage4.OSA4_6.layers.2.OSA4_6_
2/norm.running_var", "bottom_up.stage4.OSA4_6.layers.3.OSA4_6_3/conv.weight", "bottom_up.stage4.OSA4_6.layers.3.OSA4_6_3/norm.weight", "bottom_up.stage4.OSA4_6.layers.3
.OSA4_6_3/norm.bias", "bottom_up.stage4.OSA4_6.layers.3.OSA4_6_3/norm.running_mean", "bottom_up.stage4.OSA4_6.layers.3.OSA4_6_3/norm.running_var", "bottom_up.stage4.OSA
4_6.layers.4.OSA4_6_4/conv.weight", "bottom_up.stage4.OSA4_6.layers.4.OSA4_6_4/norm.weight", "bottom_up.stage4.OSA4_6.layers.4.OSA4_6_4/norm.bias", "bottom_up.stage4.OS
A4_6.layers.4.OSA4_6_4/norm.running_mean", "bottom_up.stage4.OSA4_6.layers.4.OSA4_6_4/norm.running_var", "bottom_up.stage4.OSA4_6.concat.OSA4_6_concat/conv.weight", "bo
ttom_up.stage4.OSA4_6.concat.OSA4_6_concat/norm.weight", "bottom_up.stage4.OSA4_6.concat.OSA4_6_concat/norm.bias", "bottom_up.stage4.OSA4_6.concat.OSA4_6_concat/norm.ru
nning_mean", "bottom_up.stage4.OSA4_6.concat.OSA4_6_concat/norm.running_var", "bottom_up.stage4.OSA4_6.ese.fc.weight", "bottom_up.stage4.OSA4_6.ese.fc.bias", "bottom_up
.stage4.OSA4_7.layers.0.OSA4_7_0/conv.weight", "bottom_up.stage4.OSA4_7.layers.0.OSA4_7_0/norm.weight", "bottom_up.stage4.OSA4_7.layers.0.OSA4_7_0/norm.bias", "bottom_u
p.stage4.OSA4_7.layers.0.OSA4_7_0/norm.running_mean", "bottom_up.stage4.OSA4_7.layers.0.OSA4_7_0/norm.running_var", "bottom_up.stage4.OSA4_7.layers.1.OSA4_7_1/conv.weig
ht", "bottom_up.stage4.OSA4_7.layers.1.OSA4_7_1/norm.weight", "bottom_up.stage4.OSA4_7.layers.1.OSA4_7_1/norm.bias", "bottom_up.stage4.OSA4_7.layers.1.OSA4_7_1/norm.run
ning_mean", "bottom_up.stage4.OSA4_7.layers.1.OSA4_7_1/norm.running_var", "bottom_up.stage4.OSA4_7.layers.2.OSA4_7_2/conv.weight", "bottom_up.stage4.OSA4_7.layers.2.OSA
4_7_2/norm.weight", "bottom_up.stage4.OSA4_7.layers.2.OSA4_7_2/norm.bias", "bottom_up.stage4.OSA4_7.layers.2.OSA4_7_2/norm.running_mean", "bottom_up.stage4.OSA4_7.layer
s.2.OSA4_7_2/norm.running_var", "bottom_up.stage4.OSA4_7.layers.3.OSA4_7_3/conv.weight", "bottom_up.stage4.OSA4_7.layers.3.OSA4_7_3/norm.weight", "bottom_up.stage4.OSA4
_7.layers.3.OSA4_7_3/norm.bias", "bottom_up.stage4.OSA4_7.layers.3.OSA4_7_3/norm.running_mean", "bottom_up.stage4.OSA4_7.layers.3.OSA4_7_3/norm.running_var", "bottom_up
.stage4.OSA4_7.layers.4.OSA4_7_4/conv.weight", "bottom_up.stage4.OSA4_7.layers.4.OSA4_7_4/norm.weight", "bottom_up.stage4.OSA4_7.layers.4.OSA4_7_4/norm.bias", "bottom_u
p.stage4.OSA4_7.layers.4.OSA4_7_4/norm.running_mean", "bottom_up.stage4.OSA4_7.layers.4.OSA4_7_4/norm.running_var", "bottom_up.stage4.OSA4_7.concat.OSA4_7_concat/conv.w
eight", "bottom_up.stage4.OSA4_7.concat.OSA4_7_concat/norm.weight", "bottom_up.stage4.OSA4_7.concat.OSA4_7_concat/norm.bias", "bottom_up.stage4.OSA4_7.concat.OSA4_7_con
cat/norm.running_mean", "bottom_up.stage4.OSA4_7.concat.OSA4_7_concat/norm.running_var", "bottom_up.stage4.OSA4_7.ese.fc.weight", "bottom_up.stage4.OSA4_7.ese.fc.bias",
 "bottom_up.stage4.OSA4_8.layers.0.OSA4_8_0/conv.weight", "bottom_up.stage4.OSA4_8.layers.0.OSA4_8_0/norm.weight", "bottom_up.stage4.OSA4_8.layers.0.OSA4_8_0/norm.bias"
, "bottom_up.stage4.OSA4_8.layers.0.OSA4_8_0/norm.running_mean", "bottom_up.stage4.OSA4_8.layers.0.OSA4_8_0/norm.running_var", "bottom_up.stage4.OSA4_8.layers.1.OSA4_8_
1/conv.weight", "bottom_up.stage4.OSA4_8.layers.1.OSA4_8_1/norm.weight", "bottom_up.stage4.OSA4_8.layers.1.OSA4_8_1/norm.bias", "bottom_up.stage4.OSA4_8.layers.1.OSA4_8
_1/norm.running_mean", "bottom_up.stage4.OSA4_8.layers.1.OSA4_8_1/norm.running_var", "bottom_up.stage4.OSA4_8.layers.2.OSA4_8_2/conv.weight", "bottom_up.stage4.OSA4_8.l
ayers.2.OSA4_8_2/norm.weight", "bottom_up.stage4.OSA4_8.layers.2.OSA4_8_2/norm.bias", "bottom_up.stage4.OSA4_8.layers.2.OSA4_8_2/norm.running_mean", "bottom_up.stage4.O
SA4_8.layers.2.OSA4_8_2/norm.running_var", "bottom_up.stage4.OSA4_8.layers.3.OSA4_8_3/conv.weight", "bottom_up.stage4.OSA4_8.layers.3.OSA4_8_3/norm.weight", "bottom_up.
stage4.OSA4_8.layers.3.OSA4_8_3/norm.bias", "bottom_up.stage4.OSA4_8.layers.3.OSA4_8_3/norm.running_mean", "bottom_up.stage4.OSA4_8.layers.3.OSA4_8_3/norm.running_var",
 "bottom_up.stage4.OSA4_8.layers.4.OSA4_8_4/conv.weight", "bottom_up.stage4.OSA4_8.layers.4.OSA4_8_4/norm.weight", "bottom_up.stage4.OSA4_8.layers.4.OSA4_8_4/norm.bias"
, "bottom_up.stage4.OSA4_8.layers.4.OSA4_8_4/norm.running_mean", "bottom_up.stage4.OSA4_8.layers.4.OSA4_8_4/norm.running_var", "bottom_up.stage4.OSA4_8.concat.OSA4_8_concat/conv.weight", "bottom_up.stage4.OSA4_8.concat.OSA4_8_concat/norm.weight", "bottom_up.stage4.OSA4_8.concat.OSA4_8_concat/norm.bias", "bottom_up.stage4.OSA4_8.concat
.OSA4_8_concat/norm.running_mean", "bottom_up.stage4.OSA4_8.concat.OSA4_8_concat/norm.running_var", "bottom_up.stage4.OSA4_8.ese.fc.weight", "bottom_up.stage4.OSA4_8.es
e.fc.bias", "bottom_up.stage4.OSA4_9.layers.0.OSA4_9_0/conv.weight", "bottom_up.stage4.OSA4_9.layers.0.OSA4_9_0/norm.weight", "bottom_up.stage4.OSA4_9.layers.0.OSA4_9_0
/norm.bias", "bottom_up.stage4.OSA4_9.layers.0.OSA4_9_0/norm.running_mean", "bottom_up.stage4.OSA4_9.layers.0.OSA4_9_0/norm.running_var", "bottom_up.stage4.OSA4_9.layer
s.1.OSA4_9_1/conv.weight", "bottom_up.stage4.OSA4_9.layers.1.OSA4_9_1/norm.weight", "bottom_up.stage4.OSA4_9.layers.1.OSA4_9_1/norm.bias", "bottom_up.stage4.OSA4_9.laye
rs.1.OSA4_9_1/norm.running_mean", "bottom_up.stage4.OSA4_9.layers.1.OSA4_9_1/norm.running_var", "bottom_up.stage4.OSA4_9.layers.2.OSA4_9_2/conv.weight", "bottom_up.stag
e4.OSA4_9.layers.2.OSA4_9_2/norm.weight", "bottom_up.stage4.OSA4_9.layers.2.OSA4_9_2/norm.bias", "bottom_up.stage4.OSA4_9.layers.2.OSA4_9_2/norm.running_mean", "bottom_
up.stage4.OSA4_9.layers.2.OSA4_9_2/norm.running_var", "bottom_up.stage4.OSA4_9.layers.3.OSA4_9_3/conv.weight", "bottom_up.stage4.OSA4_9.layers.3.OSA4_9_3/norm.weight", 
"bottom_up.stage4.OSA4_9.layers.3.OSA4_9_3/norm.bias", "bottom_up.stage4.OSA4_9.layers.3.OSA4_9_3/norm.running_mean", "bottom_up.stage4.OSA4_9.layers.3.OSA4_9_3/norm.ru
nning_var", "bottom_up.stage4.OSA4_9.layers.4.OSA4_9_4/conv.weight", "bottom_up.stage4.OSA4_9.layers.4.OSA4_9_4/norm.weight", "bottom_up.stage4.OSA4_9.layers.4.OSA4_9_4
/norm.bias", "bottom_up.stage4.OSA4_9.layers.4.OSA4_9_4/norm.running_mean", "bottom_up.stage4.OSA4_9.layers.4.OSA4_9_4/norm.running_var", "bottom_up.stage4.OSA4_9.conca
t.OSA4_9_concat/conv.weight", "bottom_up.stage4.OSA4_9.concat.OSA4_9_concat/norm.weight", "bottom_up.stage4.OSA4_9.concat.OSA4_9_concat/norm.bias", "bottom_up.stage4.OS
A4_9.concat.OSA4_9_concat/norm.running_mean", "bottom_up.stage4.OSA4_9.concat.OSA4_9_concat/norm.running_var", "bottom_up.stage4.OSA4_9.ese.fc.weight", "bottom_up.stage
4.OSA4_9.ese.fc.bias", "bottom_up.stage5.OSA5_1.layers.0.OSA5_1_0/conv.weight", "bottom_up.stage5.OSA5_1.layers.0.OSA5_1_0/norm.weight", "bottom_up.stage5.OSA5_1.layers
.0.OSA5_1_0/norm.bias", "bottom_up.stage5.OSA5_1.layers.0.OSA5_1_0/norm.running_mean", "bottom_up.stage5.OSA5_1.layers.0.OSA5_1_0/norm.running_var", "bottom_up.stage5.O
SA5_1.layers.1.OSA5_1_1/conv.weight", "bottom_up.stage5.OSA5_1.layers.1.OSA5_1_1/norm.weight", "bottom_up.stage5.OSA5_1.layers.1.OSA5_1_1/norm.bias", "bottom_up.stage5.
OSA5_1.layers.1.OSA5_1_1/norm.running_mean", "bottom_up.stage5.OSA5_1.layers.1.OSA5_1_1/norm.running_var", "bottom_up.stage5.OSA5_1.layers.2.OSA5_1_2/conv.weight", "bot
tom_up.stage5.OSA5_1.layers.2.OSA5_1_2/norm.weight", "bottom_up.stage5.OSA5_1.layers.2.OSA5_1_2/norm.bias", "bottom_up.stage5.OSA5_1.layers.2.OSA5_1_2/norm.running_mean
", "bottom_up.stage5.OSA5_1.layers.2.OSA5_1_2/norm.running_var", "bottom_up.stage5.OSA5_1.layers.3.OSA5_1_3/conv.weight", "bottom_up.stage5.OSA5_1.layers.3.OSA5_1_3/nor
m.weight", "bottom_up.stage5.OSA5_1.layers.3.OSA5_1_3/norm.bias", "bottom_up.stage5.OSA5_1.layers.3.OSA5_1_3/norm.running_mean", "bottom_up.stage5.OSA5_1.layers.3.OSA5_
1_3/norm.running_var", "bottom_up.stage5.OSA5_1.layers.4.OSA5_1_4/conv.weight", "bottom_up.stage5.OSA5_1.layers.4.OSA5_1_4/norm.weight", "bottom_up.stage5.OSA5_1.layers
.4.OSA5_1_4/norm.bias", "bottom_up.stage5.OSA5_1.layers.4.OSA5_1_4/norm.running_mean", "bottom_up.stage5.OSA5_1.layers.4.OSA5_1_4/norm.running_var", "bottom_up.stage5.O
SA5_1.concat.OSA5_1_concat/conv.weight", "bottom_up.stage5.OSA5_1.concat.OSA5_1_concat/norm.weight", "bottom_up.stage5.OSA5_1.concat.OSA5_1_concat/norm.bias", "bottom_u
p.stage5.OSA5_1.concat.OSA5_1_concat/norm.running_mean", "bottom_up.stage5.OSA5_1.concat.OSA5_1_concat/norm.running_var", "bottom_up.stage5.OSA5_1.ese.fc.weight", "bott
om_up.stage5.OSA5_1.ese.fc.bias", "bottom_up.stage5.OSA5_2.layers.0.OSA5_2_0/conv.weight", "bottom_up.stage5.OSA5_2.layers.0.OSA5_2_0/norm.weight", "bottom_up.stage5.OS
A5_2.layers.0.OSA5_2_0/norm.bias", "bottom_up.stage5.OSA5_2.layers.0.OSA5_2_0/norm.running_mean", "bottom_up.stage5.OSA5_2.layers.0.OSA5_2_0/norm.running_var", "bottom_
up.stage5.OSA5_2.layers.1.OSA5_2_1/conv.weight", "bottom_up.stage5.OSA5_2.layers.1.OSA5_2_1/norm.weight", "bottom_up.stage5.OSA5_2.layers.1.OSA5_2_1/norm.bias", "bottom
_up.stage5.OSA5_2.layers.1.OSA5_2_1/norm.running_mean", "bottom_up.stage5.OSA5_2.layers.1.OSA5_2_1/norm.running_var", "bottom_up.stage5.OSA5_2.layers.2.OSA5_2_2/conv.we
ight", "bottom_up.stage5.OSA5_2.layers.2.OSA5_2_2/norm.weight", "bottom_up.stage5.OSA5_2.layers.2.OSA5_2_2/norm.bias", "bottom_up.stage5.OSA5_2.layers.2.OSA5_2_2/norm.r
unning_mean", "bottom_up.stage5.OSA5_2.layers.2.OSA5_2_2/norm.running_var", "bottom_up.stage5.OSA5_2.layers.3.OSA5_2_3/conv.weight", "bottom_up.stage5.OSA5_2.layers.3.O
SA5_2_3/norm.weight", "bottom_up.stage5.OSA5_2.layers.3.OSA5_2_3/norm.bias", "bottom_up.stage5.OSA5_2.layers.3.OSA5_2_3/norm.running_mean", "bottom_up.stage5.OSA5_2.lay
ers.3.OSA5_2_3/norm.running_var", "bottom_up.stage5.OSA5_2.layers.4.OSA5_2_4/conv.weight", "bottom_up.stage5.OSA5_2.layers.4.OSA5_2_4/norm.weight", "bottom_up.stage5.OS
A5_2.layers.4.OSA5_2_4/norm.bias", "bottom_up.stage5.OSA5_2.layers.4.OSA5_2_4/norm.running_mean", "bottom_up.stage5.OSA5_2.layers.4.OSA5_2_4/norm.running_var", "bottom_
up.stage5.OSA5_2.concat.OSA5_2_concat/conv.weight", "bottom_up.stage5.OSA5_2.concat.OSA5_2_concat/norm.weight", "bottom_up.stage5.OSA5_2.concat.OSA5_2_concat/norm.bias"
, "bottom_up.stage5.OSA5_2.concat.OSA5_2_concat/norm.running_mean", "bottom_up.stage5.OSA5_2.concat.OSA5_2_concat/norm.running_var", "bottom_up.stage5.OSA5_2.ese.fc.wei
ght", "bottom_up.stage5.OSA5_2.ese.fc.bias", "bottom_up.stage5.OSA5_3.layers.0.OSA5_3_0/conv.weight", "bottom_up.stage5.OSA5_3.layers.0.OSA5_3_0/norm.weight", "bottom_u
p.stage5.OSA5_3.layers.0.OSA5_3_0/norm.bias", "bottom_up.stage5.OSA5_3.layers.0.OSA5_3_0/norm.running_mean", "bottom_up.stage5.OSA5_3.layers.0.OSA5_3_0/norm.running_var
", "bottom_up.stage5.OSA5_3.layers.1.OSA5_3_1/conv.weight", "bottom_up.stage5.OSA5_3.layers.1.OSA5_3_1/norm.weight", "bottom_up.stage5.OSA5_3.layers.1.OSA5_3_1/norm.bia
s", "bottom_up.stage5.OSA5_3.layers.1.OSA5_3_1/norm.running_mean", "bottom_up.stage5.OSA5_3.layers.1.OSA5_3_1/norm.running_var", "bottom_up.stage5.OSA5_3.layers.2.OSA5_
3_2/conv.weight", "bottom_up.stage5.OSA5_3.layers.2.OSA5_3_2/norm.weight", "bottom_up.stage5.OSA5_3.layers.2.OSA5_3_2/norm.bias", "bottom_up.stage5.OSA5_3.layers.2.OSA5
_3_2/norm.running_mean", "bottom_up.stage5.OSA5_3.layers.2.OSA5_3_2/norm.running_var", "bottom_up.stage5.OSA5_3.layers.3.OSA5_3_3/conv.weight", "bottom_up.stage5.OSA5_3
.layers.3.OSA5_3_3/norm.weight", "bottom_up.stage5.OSA5_3.layers.3.OSA5_3_3/norm.bias", "bottom_up.stage5.OSA5_3.layers.3.OSA5_3_3/norm.running_mean", "bottom_up.stage5
.OSA5_3.layers.3.OSA5_3_3/norm.running_var", "bottom_up.stage5.OSA5_3.layers.4.OSA5_3_4/conv.weight", "bottom_up.stage5.OSA5_3.layers.4.OSA5_3_4/norm.weight", "bottom_u
p.stage5.OSA5_3.layers.4.OSA5_3_4/norm.bias", "bottom_up.stage5.OSA5_3.layers.4.OSA5_3_4/norm.running_mean", "bottom_up.stage5.OSA5_3.layers.4.OSA5_3_4/norm.running_var
", "bottom_up.stage5.OSA5_3.concat.OSA5_3_concat/conv.weight", "bottom_up.stage5.OSA5_3.concat.OSA5_3_concat/norm.weight", "bottom_up.stage5.OSA5_3.concat.OSA5_3_concat
/norm.bias", "bottom_up.stage5.OSA5_3.concat.OSA5_3_concat/norm.running_mean", "bottom_up.stage5.OSA5_3.concat.OSA5_3_concat/norm.running_var", "bottom_up.stage5.OSA5_3
.ese.fc.weight", "bottom_up.stage5.OSA5_3.ese.fc.bias".                                                                                                                 
        Unexpected key(s) in state_dict: "model", "optimizer", "scheduler", "iteration".

I'd appreciate some help.

Can release DD3DV2 code？

Hi，can release DD3DV2 code? Thank you!

I can't reproduce the result of Kitti Dataset.

I can't reproduce the result of Kitti Dataset. By comparing the official log and my training log, there are some differences in tht config file, for example. Can anyone answer this question?

Question about the releasing time

Hello, @dennis-park-TRI !
Thanks for your excellent work DD3D, which shows great results in benchmark. I am wondering when will you open-source your code.

Multi-node training

Hi there,
Thank you so much for this release!
When trying to run multi-node training, I can see that this repo is equipped to do this, when I see the following lines:

dd3d/tridet/utils/setup.py

Line 57 in da25b61

# Multi-node training often fails with "received 0 items of ancdata" error.

dd3d/Makefile

Line 42 in da25b61

-H ${MPI_HOSTS} \

Have you trained using multiple nodes (not just multiple GPUs) where you have to provide 2 different ip addresses from within the docker containers you provided in this repo? And has this worked for you? When I execute training on two different machines, the code hangs and I dont see any terminal printouts...

Thank you in advance!

Guidance on Pre training

Hello，thanks for sharing your excellent work!
I would like to adapt your work to other model, for example, use resnet50 as backbone, and refactor head branches.
Could your please share the pretrain codes or give me some advice.
Thanks!

Unable to get visualize_dataloader.py to run due to kitti access

I'm getting the following error when I run the visualize_dataloader.py file in the instructions. I verified that I have the file it's looking for in the directory. I believe there is a problem with pulling in the root directory as it's looking for /data/datasets directly from the /dd3d/ directory level.

[09/12 20:48:50 tridet.data.datasets.kitti_3d]: KITTI-3D dataset(s): kitti_3d_train, kitti_3d_val 
Error executing job with overrides: ['+experiments=dd3d_kitti_dla34', 'SOLVER.IMS_PER_BATCH=4']
Traceback (most recent call last):
  File "./scripts/visualize_dataloader.py", line 26, in main
    dataset_names = register_datasets(cfg)
  File "/workspace/dd3d/tridet/data/datasets/__init__.py", line 19, in register_datasets
    dataset_names.extend(register_kitti_3d_datasets(required_datasets, cfg))
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/__init__.py", line 41, in register_kitti_3d_datasets
    fn(name, **kwargs)
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 312, in register_kitti_3d_metadata
    dataset_dicts = DatasetCatalog.get(dataset_name)
  File "/usr/local/lib/python3.8/dist-packages/detectron2/data/catalog.py", line 58, in get
    return f()
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 298, in build_monocular_kitti3d_dataset
    dataset = KITTI3DMonocularDataset(root_dir, mv3d_split, class_names, sensors, box2d_from_box3d, max_num_items)
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 283, in __init__
    self._kitti_dset = KITTI3DDataset(root_dir, mv3d_split, class_names, sensors, box2d_from_box3d, max_num_items)
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 57, in __init__
    with open(os.path.join(self.root_dir, "mv3d_kitti_splits", "{}.txt".format(mv3d_split))) as _f:
FileNotFoundError: [Errno 2] No such file or directory: '/data/datasets/KITTI3D/mv3d_kitti_splits/train.txt'

how to train DD3D on custom datasets

Hello,

I only see the KITTI and Nuscense dataloader. Could you tell me how to build custom datasets which we can use to train DD3D.
Or should we try to convert our custom dataset into KITTI or Nuscense format first, and then use their corresponding dataloader?

I am looking forward to your reply.

Thanks

Omninets weights Some model parameters or buffers are not found

[03/17 17:31:59 tridet]: Registered 2 datasets:
kitti_3d_train
kitti_3d_val
[03/17 17:31:59 fvcore.common.checkpoint]: [Checkpointer] Loading from /home/azuryl/dd3d_test/model/depth_pretrained_omninet-small-3nxjur71.pth ...
WARNING [03/17 17:31:59 fvcore.common.checkpoint]: Some model parameters or buffers are not found in the checkpoint:
fcos2d_head.box2d_reg.{bias, weight}
fcos2d_head.box2d_tower.0.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.0.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.0.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.0.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.0.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.0.weight
fcos2d_head.box2d_tower.1.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.1.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.1.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.1.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.1.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.1.weight
fcos2d_head.box2d_tower.2.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.2.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.2.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.2.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.2.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.2.weight
fcos2d_head.box2d_tower.3.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.3.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.3.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.3.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.3.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.box2d_tower.3.weight
fcos2d_head.centerness.{bias, weight}
fcos2d_head.cls_logits.{bias, weight}
fcos2d_head.cls_tower.0.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.0.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.0.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.0.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.0.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.0.weight
fcos2d_head.cls_tower.1.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.1.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.1.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.1.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.1.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.1.weight
fcos2d_head.cls_tower.2.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.2.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.2.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.2.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.2.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.2.weight
fcos2d_head.cls_tower.3.norm.0.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.3.norm.1.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.3.norm.2.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.3.norm.3.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.3.norm.4.{bias, running_mean, running_var, weight}
fcos2d_head.cls_tower.3.weight
fcos2d_head.scales_box2d_reg.0.scale
fcos2d_head.scales_box2d_reg.1.scale
fcos2d_head.scales_box2d_reg.2.scale
fcos2d_head.scales_box2d_reg.3.scale
fcos2d_head.scales_box2d_reg.4.scale
fcos3d_head.box3d_conf.0.{bias, weight}
fcos3d_head.box3d_ctr.0.{bias, weight}
fcos3d_head.box3d_depth.0.weight
fcos3d_head.box3d_quat.0.{bias, weight}
fcos3d_head.box3d_size.0.{bias, weight}
fcos3d_head.box3d_tower.0.norm.0.{bias, weight}
fcos3d_head.box3d_tower.0.norm.1.{bias, weight}
fcos3d_head.box3d_tower.0.norm.2.{bias, weight}
fcos3d_head.box3d_tower.0.norm.3.{bias, weight}
fcos3d_head.box3d_tower.0.norm.4.{bias, weight}
fcos3d_head.box3d_tower.0.weight
fcos3d_head.box3d_tower.1.norm.0.{bias, weight}
fcos3d_head.box3d_tower.1.norm.1.{bias, weight}
fcos3d_head.box3d_tower.1.norm.2.{bias, weight}
fcos3d_head.box3d_tower.1.norm.3.{bias, weight}
fcos3d_head.box3d_tower.1.norm.4.{bias, weight}
fcos3d_head.box3d_tower.1.weight
fcos3d_head.box3d_tower.2.norm.0.{bias, weight}
fcos3d_head.box3d_tower.2.norm.1.{bias, weight}
fcos3d_head.box3d_tower.2.norm.2.{bias, weight}
fcos3d_head.box3d_tower.2.norm.3.{bias, weight}
fcos3d_head.box3d_tower.2.norm.4.{bias, weight}
fcos3d_head.box3d_tower.2.weight
fcos3d_head.box3d_tower.3.norm.0.{bias, weight}
fcos3d_head.box3d_tower.3.norm.1.{bias, weight}
fcos3d_head.box3d_tower.3.norm.2.{bias, weight}
fcos3d_head.box3d_tower.3.norm.3.{bias, weight}
fcos3d_head.box3d_tower.3.norm.4.{bias, weight}
fcos3d_head.box3d_tower.3.weight
fcos3d_head.offsets_depth.0.bias
fcos3d_head.offsets_depth.1.bias
fcos3d_head.offsets_depth.2.bias
fcos3d_head.offsets_depth.3.bias
fcos3d_head.offsets_depth.4.bias
fcos3d_head.scales_conf.0.scale
fcos3d_head.scales_conf.1.scale
fcos3d_head.scales_conf.2.scale
fcos3d_head.scales_conf.3.scale
fcos3d_head.scales_conf.4.scale
fcos3d_head.scales_depth.0.scale
fcos3d_head.scales_depth.1.scale
fcos3d_head.scales_depth.2.scale
fcos3d_head.scales_depth.3.scale
fcos3d_head.scales_depth.4.scale
fcos3d_head.scales_proj_ctr.0.scale
fcos3d_head.scales_proj_ctr.1.scale
fcos3d_head.scales_proj_ctr.2.scale
fcos3d_head.scales_proj_ctr.3.scale
fcos3d_head.scales_proj_ctr.4.scale
fcos3d_head.scales_size.0.scale
fcos3d_head.scales_size.1.scale
fcos3d_head.scales_size.2.scale
fcos3d_head.scales_size.3.scale
fcos3d_head.scales_size.4.scale
fcos3d_head.{mean_depth_per_level, std_depth_per_level}
pixel_mean
pixel_std
WARNING [03/17 17:31:59 fvcore.common.checkpoint]: The checkpoint state_dict contains keys that are not used by the model:
depth_head.net.conv_layers.0.weight
depth_head.net.conv_layers.1.weight
depth_head.net.conv_layers.2.weight
depth_head.net.conv_layers.3.weight
depth_head.net.bn_layers.0.0.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.0.1.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.0.2.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.0.3.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.1.0.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.1.1.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.1.2.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.1.3.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.2.0.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.2.1.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.2.2.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.2.3.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.3.0.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.3.1.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.3.2.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.3.3.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.4.0.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.4.1.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.4.2.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.net.bn_layers.4.3.{bias, num_batches_tracked, running_mean, running_var, weight}
depth_head.predictor.{bias, weight}

Interesting

Big progress, good architecture. The inference latency should also be reduced.

But I have two questions.
What's the density of 3d points outputted by the feature map? [Should as dense as image resolution, It doesn't matter.]
As we know, image is more dense compared to point cloud produced by lidar, even dense lidar. [Alright, we can sub-sample it latter.]
OK, state of the art 3d detector head can be used, such as center-point, it is efficient.

Hum, I think the only latency bottleneck may be the dense 3d point sub-sample part.
3d points sub-sample methods designed for Lidar generated data is in-efficient here, should be the bottleneck.
The output of the confidence could be used here. But, is the predicted confidence confidence?
Hope to see someone help to design good sub-sample method here.

Right, I have not read the code. Just a glimps of the paper. Lots of self guess.

model trained using nuscense dataset

Hi, @dennis-park-TRI , Thanks for your code release.
I am wondering when will the model trained using nuscense dataset be released?

Can you please share the coco pretrained model?

Hi TRI team,

Thanks for sharing this exellent work! I have lately been running this code and want to run some experiments from the scratch. Could you please share your DD3D model pretrained ONLY on the COCO dataset (without depth pretraining)?

And also, just want to be clear, the model you have provided are pretrained on COCO AND DDAD15M, right?

Thank you and I am looking forward to your reply!

Regards,
Johan

'ValueError: cannot reshape array of size 14 into shape (4)' when running scripts

Hi,

Whenever I run evaluation on sample of the Kitti dataset. I get this error. I also get the same error when
running the following script:
./scripts/visualize_dataloader.py +experiments=dd3d_kitti_dla34 SOLVER.IMS_PER_BATCH=4
Here are my terminal logs from running the above command:

No protocol specified
No protocol specified
No protocol specified
/usr/local/lib/python3.8/dist-packages/hydra/_internal/defaults_list.py:251: UserWarning: In 'visualize_dataloader': Defaults list is missing `_self_`. See https://hydra.cc/docs/upgrades/1.0_to_1.1/default_composition_order for more information
  warnings.warn(msg, UserWarning)
[11/02 13:22:33 tridet.utils.s3]: Downloading initial weights:
[11/02 13:22:33 tridet.utils.s3]:   src: https://tri-ml-public.s3.amazonaws.com/github/dd3d/pretrained/depth_pretrained_dla34-y1urdmir-20210422_165446-model_final-remapped.pth
[11/02 13:22:33 tridet.utils.s3]:   dst: /tmp/tmpwxys0idg.pth
835it [00:00, 25902.64it/s]
[11/02 13:28:08 tridet.utils.hydra.callbacks]: Rank of current process: 0. World size: 1
[11/02 13:28:08 tridet.utils.setup]: Working Directory: /workspace/dd3d/outputs/2021-11-02/13-22-33
[11/02 13:28:08 tridet.utils.setup]: Full config:
{
  "WANDB": {
    "ENABLED": false,
    "DRYRUN": false,
    "PROJECT": "dd3d",
    "GROUP": null,
    "TAGS": [
      "kitti-val",
      "dla34",
      "bn"
    ]
  },
  "EVAL_ONLY": false,
  "EVAL_ON_START": false,
  "ONLY_REGISTER_DATASETS": false,
  "OUTPUT_ROOT": "./outputs",
  "SYNC_OUTPUT_DIR_S3": {
    "ENABLED": false,
    "ROOT_IN_S3": "???",
    "PERIOD": 1000
  },
  "DATASET_ROOT": "/data/datasets/",
  "TMP_DIR": "/tmp/",
  "DATASETS": {
    "TRAIN": {
      "NAME": "kitti_3d_train",
      "CANONICAL_BOX3D_SIZES": [
        [
          1.61876949,
          3.89154523,
          1.52969237
        ],
        [
          0.62806586,
          0.82038497,
          1.76784787
        ],
        [
          0.56898187,
          1.77149234,
          1.7237099
        ],
        [
          1.9134491,
          5.15499603,
          2.18998422
        ],
        [
          2.61168401,
          9.22692319,
          3.36492722
        ],
        [
          0.5390196,
          1.08098042,
          1.28392158
        ],
        [
          2.36044838,
          15.56991038,
          3.5289238
        ],
        [
          1.24489164,
          2.51495357,
          1.61402478
        ]
      ],
      "DATASET_MAPPER": "default",
      "NUM_CLASSES": 5,
      "MEAN_DEPTH_PER_LEVEL": [
        32.594,
        15.178,
        8.424,
        5.004,
        4.662
      ],
      "STD_DEPTH_PER_LEVEL": [
        14.682,
        7.139,
        4.345,
        2.399,
        2.587
      ]
    },
    "TEST": {
      "NAME": "kitti_3d_val",
      "NUSC_SAMPLE_AGGREGATE_IN_INFERENCE": false,
      "DATASET_MAPPER": "default"
    }
  },
  "FE": {
    "FPN": {
      "IN_FEATURES": [
        "level3",
        "level4",
        "level5"
      ],
      "OUT_FEATURES": null,
      "OUT_CHANNELS": 256,
      "NORM": "FrozenBN",
      "FUSE_TYPE": "sum"
    },
    "BUILDER": "build_fcos_dla_fpn_backbone_p67",
    "BACKBONE": {
      "NAME": "DLA-34",
      "OUT_FEATURES": [
        "level3",
        "level4",
        "level5"
      ],
      "NORM": "FrozenBN"
    },
    "OUT_FEATURES": null
  },
  "DD3D": {
    "IN_FEATURES": null,
    "NUM_CLASSES": 5,
    "FEATURE_LOCATIONS_OFFSET": "none",
    "SIZES_OF_INTEREST": [
      64,
      128,
      256,
      512
    ],
    "INFERENCE": {
      "DO_NMS": true,
      "DO_POSTPROCESS": true,
      "DO_BEV_NMS": false,
      "BEV_NMS_IOU_THRESH": 0.3,
      "NUSC_SAMPLE_AGGREGATE": false
    },
    "FCOS2D": {
      "_VERSION": "v2",
      "NORM": "BN",
      "NUM_CLS_CONVS": 4,
      "NUM_BOX_CONVS": 4,
      "USE_DEFORMABLE": false,
      "USE_SCALE": true,
      "BOX2D_SCALE_INIT_FACTOR": 1.0,
      "LOSS": {
        "ALPHA": 0.25,
        "GAMMA": 2.0,
        "LOC_LOSS_TYPE": "giou"
      },
      "INFERENCE": {
        "THRESH_WITH_CTR": true,
        "PRE_NMS_THRESH": 0.05,
        "PRE_NMS_TOPK": 1000,
        "POST_NMS_TOPK": 100,
        "NMS_THRESH": 0.75
      }
    },
    "FCOS3D": {
      "NORM": "FrozenBN",
      "NUM_CONVS": 4,
      "USE_DEFORMABLE": false,
      "USE_SCALE": true,
      "DEPTH_SCALE_INIT_FACTOR": 0.3,
      "PROJ_CTR_SCALE_INIT_FACTOR": 1.0,
      "PER_LEVEL_PREDICTORS": false,
      "SCALE_DEPTH_BY_FOCAL_LENGTHS": true,
      "SCALE_DEPTH_BY_FOCAL_LENGTHS_FACTOR": 500.0,
      "MEAN_DEPTH_PER_LEVEL": [
        32.594,
        15.178,
        8.424,
        5.004,
        4.662
      ],
      "STD_DEPTH_PER_LEVEL": [
        14.682,
        7.139,
        4.345,
        2.399,
        2.587
      ],
      "MIN_DEPTH": 0.1,
      "MAX_DEPTH": 80.0,
      "CANONICAL_BOX3D_SIZES": [
        [
          1.61876949,
          3.89154523,
          1.52969237
        ],
        [
          0.62806586,
          0.82038497,
          1.76784787
        ],
        [
          0.56898187,
          1.77149234,
          1.7237099
        ],
        [
          1.9134491,
          5.15499603,
          2.18998422
        ],
        [
          2.61168401,
          9.22692319,
          3.36492722
        ],
        [
          0.5390196,
          1.08098042,
          1.28392158
        ],
        [
          2.36044838,
          15.56991038,
          3.5289238
        ],
        [
          1.24489164,
          2.51495357,
          1.61402478
        ]
      ],
      "CLASS_AGNOSTIC_BOX3D": false,
      "PREDICT_ALLOCENTRIC_ROT": true,
      "PREDICT_DISTANCE": false,
      "LOSS": {
        "SMOOTH_L1_BETA": 0.05,
        "MAX_LOSS_PER_GROUP_DISENT": 20.0,
        "CONF_3D_TEMPERATURE": 1.0,
        "WEIGHT_BOX3D": 2.0,
        "WEIGHT_CONF3D": 1.0
      },
      "PREPARE_TARGET": {
        "CENTER_SAMPLE": true,
        "POS_RADIUS": 1.5
      }
    }
  },
  "VIS": {
    "DATALOADER_ENABLED": true,
    "DATALOADER_PERIOD": 1000,
    "DATALOADER_MAX_NUM_SAMPLES": 10,
    "PREDICTIONS_ENABLED": true,
    "PREDICTIONS_MAX_NUM_SAMPLES": 20,
    "D2": {
      "DATALOADER": {
        "ENABLED": true,
        "SCALE": 1.0,
        "COLOR_MODE": "image"
      },
      "PREDICTIONS": {
        "ENABLED": true,
        "SCALE": 1.0,
        "COLOR_MODE": "image",
        "THRESHOLD": 0.4
      }
    },
    "BOX3D": {
      "DATALOADER": {
        "ENABLED": true,
        "SCALE": 1.0,
        "RENDER_LABELS": true
      },
      "PREDICTIONS": {
        "ENABLED": true,
        "SCALE": 1.0,
        "RENDER_LABELS": true,
        "THRESHOLD": 0.5,
        "MIN_DEPTH_CENTER": 0.0
      }
    }
  },
  "INPUT": {
    "FORMAT": "BGR",
    "AUG_ENABLED": true,
    "RESIZE": {
      "ENABLED": true,
      "MIN_SIZE_TRAIN": [
        288,
        304,
        320,
        336,
        352,
        368,
        384,
        400,
        416,
        448,
        480,
        512,
        544,
        576
      ],
      "MIN_SIZE_TRAIN_SAMPLING": "choice",
      "MAX_SIZE_TRAIN": 10000,
      "MIN_SIZE_TEST": 384,
      "MAX_SIZE_TEST": 100000
    },
    "CROP": {
      "ENABLED": false,
      "TYPE": "relative_range",
      "SIZE": [
        0.9,
        0.9
      ]
    },
    "RANDOM_FLIP": {
      "ENABLED": true,
      "HORIZONTAL": true,
      "VERTICAL": false
    },
    "COLOR_JITTER": {
      "ENABLED": true,
      "BRIGHTNESS": [
        0.2,
        0.2
      ],
      "SATURATION": [
        0.2,
        0.2
      ],
      "CONTRAST": [
        0.2,
        0.2
      ]
    }
  },
  "MODEL": {
    "DEVICE": "cuda",
    "META_ARCHITECTURE": "DD3D",
    "PIXEL_MEAN": [
      103.53,
      116.28,
      123.675
    ],
    "PIXEL_STD": [
      57.375,
      57.12,
      58.395
    ],
    "CKPT": "/tmp/tmpwxys0idg.pth",
    "BOX2D_ON": true,
    "BOX3D_ON": true,
    "DEPTH_ON": false,
    "CHECKPOINT": ""
  },
  "DATALOADER": {
    "TRAIN": {
      "NUM_WORKERS": 12,
      "FILTER_EMPTY_ANNOTATIONS": true,
      "SAMPLER": "RepeatFactorTrainingSampler",
      "REPEAT_THRESHOLD": 0.4,
      "ASPECT_RATIO_GROUPING": false
    },
    "TEST": {
      "NUM_WORKERS": 4,
      "SAMPLER": "InferenceSampler"
    }
  },
  "SOLVER": {
    "IMS_PER_BATCH": 4,
    "BASE_LR": 0.002,
    "MOMENTUM": 0.9,
    "NESTEROV": false,
    "WEIGHT_DECAY": 0.0001,
    "WEIGHT_DECAY_NORM": 0.0,
    "BIAS_LR_FACTOR": 1.0,
    "WEIGHT_DECAY_BIAS": 0.0001,
    "GAMMA": 0.1,
    "LR_SCHEDULER_NAME": "WarmupMultiStepLR",
    "STEPS": [
      21500,
      24000
    ],
    "WARMUP_FACTOR": 0.0001,
    "WARMUP_ITERS": 2000,
    "WARMUP_METHOD": "linear",
    "CLIP_GRADIENTS": {
      "ENABLED": false,
      "CLIP_TYPE": "value",
      "CLIP_VALUE": 1.0,
      "NORM_TYPE": 2.0
    },
    "CHECKPOINT_PERIOD": 2000,
    "MIXED_PRECISION_ENABLED": true,
    "DDP_FIND_UNUSED_PARAMETERS": false,
    "ACCUMULATE_GRAD_BATCHES": 1,
    "SYNCBN_USE_LOCAL_WORKERS": false,
    "MAX_ITER": 25000
  },
  "TEST": {
    "ENABLED": true,
    "EVAL_PERIOD": 2000,
    "EVAL_ON_START": false,
    "ADDITIONAL_EVAL_STEPS": [],
    "IMS_PER_BATCH": 80,
    "AUG": {
      "ENABLED": true,
      "MIN_SIZES": [
        320,
        384,
        448,
        512,
        576
      ],
      "MAX_SIZE": 100000,
      "FLIP": true
    }
  },
  "USE_TEST": false,
  "EVALUATORS": {
    "KITTI3D": {
      "IOU_THRESHOLDS": [
        0.5,
        0.7
      ],
      "ONLY_PREPARE_SUBMISSION": false
    }
  }
}
[11/02 13:28:08 tridet.data.datasets.kitti_3d]: KITTI-3D dataset(s): kitti_3d_train, kitti_3d_val 
Error executing job with overrides: ['+experiments=dd3d_kitti_dla34', 'SOLVER.IMS_PER_BATCH=4']
multiprocessing.pool.RemoteTraceback: 
"""
Traceback (most recent call last):
  File "/usr/lib/python3.8/multiprocessing/pool.py", line 125, in worker
    result = (True, func(*args, **kwds))
  File "/usr/lib/python3.8/multiprocessing/pool.py", line 48, in mapstar
    return list(map(*args))
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 123, in _read_calibration_file
    P_20 = calibration.loc[2].values[1:].reshape(-1, 4).astype(np.float64)
ValueError: cannot reshape array of size 14 into shape (4)
"""

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "./scripts/visualize_dataloader.py", line 26, in main
    dataset_names = register_datasets(cfg)
  File "/workspace/dd3d/tridet/data/datasets/__init__.py", line 19, in register_datasets
    dataset_names.extend(register_kitti_3d_datasets(required_datasets, cfg))
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/__init__.py", line 41, in register_kitti_3d_datasets
    fn(name, **kwargs)
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 312, in register_kitti_3d_metadata
    dataset_dicts = DatasetCatalog.get(dataset_name)
  File "/usr/local/lib/python3.8/dist-packages/detectron2/data/catalog.py", line 58, in get
    return f()
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 298, in build_monocular_kitti3d_dataset
    dataset = KITTI3DMonocularDataset(root_dir, mv3d_split, class_names, sensors, box2d_from_box3d, max_num_items)
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 283, in __init__
    self._kitti_dset = KITTI3DDataset(root_dir, mv3d_split, class_names, sensors, box2d_from_box3d, max_num_items)
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 73, in __init__
    self.calibration_table = self._parse_calibration_files()
  File "/workspace/dd3d/tridet/data/datasets/kitti_3d/build.py", line 95, in _parse_calibration_files
    (_proc.map(self._read_calibration_file, calibration_files))
  File "/usr/lib/python3.8/multiprocessing/pool.py", line 364, in map
    return self._map_async(func, iterable, mapstar, chunksize).get()
  File "/usr/lib/python3.8/multiprocessing/pool.py", line 771, in get
    raise self._value
ValueError: cannot reshape array of size 14 into shape (4)

no module named tridet

I am getting this error when I validate my installation. There is no module called tridet in pip. google search gives no such module name.
I installed trident but I am still getting that error.

do you have requirements file with version numbers of the libraries for conda?

Multi-GPU training inside docker

HI ,

Thanks for your code release.
I have a question about Multi-GPU training command.
Is it possible to train with Multi-GPU(8) inside docker?

Like:python -m torch.distributed.launch --nproc_per_node 8 train.py xxx

Multi-GPU training outside docker by using the following command is not so comfortable for server training :

make docker-run-mpi COMMAND="".

I am looking forward to your Reply.
And thanks again for your great job!

How to evaluate original data

I wanna evaluate my own data with pre-trained model(DLA34). Evaluating with KITTI dataset succesed, but I can't evaluate own data.

How to evaluate my own data? Please tell me.

Thank you!!

find a bug in image_list.py file about the functiion of from_tensors()

@dennis-park-TRI I found a bug in the project of the image_list.py about the function of from_tensors().When padding image,I think that the intrinsics of the image can make a difference(for example cx,cy).I misunderstand it or It is indeed wrong?Looking forward to your reply.Thank you...

Generating Validation Folder

For anyone who had to generate the validation folder, here's what I used.

`import os
import shutil

root = ''
with open(os.path.join(root, "mv3d_kitti_splits", "val.txt")) as _f:
lines = _f.readlines()
split = [line.rstrip("\n") for line in lines]

for sub in ['calib', 'image_2', 'label_2']:
for file in split:
if sub == 'calib' or sub == 'label_2':
file += '.txt'
else:
file += '.png'
shutil.copyfile(os.path.join(root, 'training',sub, file), os.path.join(root, 'val',sub,file))`

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.