zhengye1995 / tianchi-2019-guangdong-intelligent-identification-of-cloth-defects-rank5 Goto Github PK

View Code? Open in Web Editor NEW

399.0 7.0 143.0 24.17 MB

天池2019广东工业智造创新大赛布匹疵点检测天池水也太深了季军解决方案

Python 85.29% Shell 0.30% C++ 4.87% Cuda 9.54%

detection tianchi computer-vision

tianchi-2019-guangdong-intelligent-identification-of-cloth-defects-rank5's Introduction

My visitor

I'm Zheng Ye

🔭 I’m currently working on JD
🌱 I’m currently learning Computer vision
👯 Currently working in Beijing, China
🤔 I love Reading, Research
💬 Ask me about Deep Learning, Computer vision and my projects
📫 How to reach me: [email protected]

⚡ Fun fact: Movie, DOTA2, NBA2K

About My Github

My Skill Set

DeepLearning

Backend

DevOps

tianchi-2019-guangdong-intelligent-identification-of-cloth-defects-rank5's People

Contributors

Stargazers

Watchers

Forkers

bxhandhxb jimmyhhua wulingtian chenlongsun cookcv uestctubiao yihtion xuner1213 sanluku liangzuan1983 cqcracked cwinjet zyg11 royzon trigrass2 czqhubu clw5180 trevolan77 2017tjm 000wangbo blackwitty williamjames007 luanchenhui yuanligesheng fengyouliang a793719595 vinceyzw p517332051 huang-git123 hell-to-heaven jinzhu6 leozhibin jkoforgh dl19940602 ssjiacv lovepan1 jackstonea bigbigxing823 goodluckcjj maxteleg intjun wishinger-li cwb96 ljc0753 chentao199617 73-seventythree single430 chang111 sxlyiyiyi jeffwin2015 liuweixiaomidi youtang1993 xwyangjshb robingong browserliu csdingithub china-liweihong 12345fengce jiangbin216 buffaloiron wuterry huanleev rockymax365 xiaopang59 x-bruce lusi1990 jinzhucn bj5546 xluckywang sonwa overlordmax pofe907 lucklyzhu yinlan fangwudi yif-yang kuanglongjkl htianlang vv-dl-jump shanhedian2017 liupengyang-1008 hairlessman beyond-zw qqyouhappy wilbur-lqw seaboat2009 kingbackyang csgaobb yustaub zhxngjxnhxx jiyuxuan926 caihaocheng-caihaocheng krisandchris fateeeeee xuewengeophysics qicekeji123456 xixi323 yangsuhui michaelzhero a963082404

tianchi-2019-guangdong-intelligent-identification-of-cloth-defects-rank5's Issues

random seed

hi
Thanks for your generous sharing very much.
I find the seed of your code is random. Does it have no influence on your results?

OOM problem when training with only single gpu

python train.py ../config/fabric_defect/cascade_rcnn_r50_fpn_70e.py

Hello, I only have one GPU (RTX2080, 8GB), so I run it on train.py and it doesn't run due to OOM problems. To solve this problem, I think I have to reduce the image size(possible?) or increase the GPU.

Could you tell me if I approach wrong or if there is a way to reduce the size of the image size?
And there is an extra graphics card, RTX3080, so I think we need to run it on CUDA 11.0 and pytorch 1.7 to run it, will this code work? I'd appreciate it if you could let me know the possibility.

Or could Google Cloud Platform be an alternative? I want to know the amount of memory I need.

Thank you

problem error log

你好，初赛的数据集能分享吗，谢谢。

需要单独划分一部分验证集出来吗

您好，非常感谢您的代码。我看了一下您的配置文件cascade_rcnn_r50_fpn_70e.py中验证集设定如下：
val=dict( type=dataset_type, ann_file=data_root + 'annotations/instances_val2017.json', img_prefix=data_root + 'val2017/', pipeline=test_pipeline),
这个好像是默认的voc数据的路径，而不是布匹缺陷检测的数据。所以，您并没有设定单独的验证我集，我这样理解对吗？

标注文件对应

您好，这个标注文件里的疵点代号怎么和疵点种类对应呢，如：\u4e09\u4e1d、\u7a00\u5bc6\u6863分别是那个疵点类型呢

请问一下怎么输入6通道的图片

请教一下收敛的速度

配置：cascade_rcnn_r50_fpn_400.py
镜像：包含了 mmdetection 1.1.0 的自定义镜像，未使用 repo 中的环境安装方法
设备：1080TI

70 epoch 差不多要 20 几个小时才能够训练完，目前处于 epoch11，想请教一下每个 epoch 大致的收敛趋势，大概每个 epoch loss 下降的的量级？或者 AP0.5:0.95 的变化大小？如果能有一份训练日志就最好了。

如下为 epoch10 -> epoch11 两次 evaluation 的结果，也想请帮忙看下是否正常，感谢。

2020-08-12 06:05:36,576 - mmdet - INFO - Epoch [1][50/181]	lr: 0.01196, eta: 21:29:16, time: 6.130, data_time: 2.914, memory: 9577, loss_rpn_cls: 0.1903, loss_rpn_bbox: 0.0545, s0.loss_cls: 0.1315, s0.acc: 98.1053, s0.loss_bbox: 0.0385, s1.loss_cls: 0.0568, s1.acc: 98.3502, s1.loss_bbox: 0.0197, s2.loss_cls: 0.0255, s2.acc: 98.5345, s2.loss_bbox: 0.0073, loss: 0.5240
2020-08-12 06:10:06,062 - mmdet - INFO - Epoch [1][100/181]	lr: 0.01396, eta: 20:06:31, time: 5.389, data_time: 1.771, memory: 9577, loss_rpn_cls: 0.1203, loss_rpn_bbox: 0.0454, s0.loss_cls: 0.1173, s0.acc: 97.6698, s0.loss_bbox: 0.0497, s1.loss_cls: 0.0536, s1.acc: 97.8707, s1.loss_bbox: 0.0332, s2.loss_cls: 0.0242, s2.acc: 98.0889, s2.loss_bbox: 0.0151, loss: 0.4589
2020-08-12 06:14:48,283 - mmdet - INFO - Epoch [1][150/181]	lr: 0.01596, eta: 19:53:50, time: 5.646, data_time: 1.851, memory: 9577, loss_rpn_cls: 0.1236, loss_rpn_bbox: 0.0452, s0.loss_cls: 0.1279, s0.acc: 97.2344, s0.loss_bbox: 0.0593, s1.loss_cls: 0.0596, s1.acc: 97.3373, s1.loss_bbox: 0.0452, s2.loss_cls: 0.0275, s2.acc: 97.4865, s2.loss_bbox: 0.0233, loss: 0.5116
2020-08-12 06:31:25,766 - mmdet - INFO - Evaluating bbox...
2020-08-12 06:32:18,841 - mmdet - INFO - Epoch [1][181/181]	lr: 0.01720, bbox_mAP: 0.0840, bbox_mAP_50: 0.1520, bbox_mAP_75: 0.0760, bbox_mAP_s: 0.0000, bbox_mAP_m: 0.0180, bbox_mAP_l: 0.0980, bbox_mAP_copypaste: 0.084 0.152 0.076 0.000 0.018 0.098
2020-08-12 06:37:07,378 - mmdet - INFO - Epoch [2][50/181]	lr: 0.01920, eta: 17:09:03, time: 5.768, data_time: 2.733, memory: 9577, loss_rpn_cls: 0.1019, loss_rpn_bbox: 0.0443, s0.loss_cls: 0.1195, s0.acc: 97.0923, s0.loss_bbox: 0.0637, s1.loss_cls: 0.0569, s1.acc: 97.1745, s1.loss_bbox: 0.0537, s2.loss_cls: 0.0268, s2.acc: 97.2980, s2.loss_bbox: 0.0304, loss: 0.4972
2020-08-12 06:41:50,079 - mmdet - INFO - Epoch [2][100/181]	lr: 0.02120, eta: 17:30:20, time: 5.655, data_time: 2.605, memory: 9577, loss_rpn_cls: 0.1054, loss_rpn_bbox: 0.0432, s0.loss_cls: 0.1162, s0.acc: 97.2228, s0.loss_bbox: 0.0613, s1.loss_cls: 0.0548, s1.acc: 97.3068, s1.loss_bbox: 0.0515, s2.loss_cls: 0.0258, s2.acc: 97.4464, s2.loss_bbox: 0.0289, loss: 0.4870
2020-08-12 06:46:36,679 - mmdet - INFO - Epoch [2][150/181]	lr: 0.02320, eta: 17:46:05, time: 5.730, data_time: 2.731, memory: 9577, loss_rpn_cls: 0.0899, loss_rpn_bbox: 0.0437, s0.loss_cls: 0.1141, s0.acc: 97.1224, s0.loss_bbox: 0.0655, s1.loss_cls: 0.0548, s1.acc: 97.1829, s1.loss_bbox: 0.0560, s2.loss_cls: 0.0262, s2.acc: 97.2586, s2.loss_bbox: 0.0319, loss: 0.4823
2020-08-12 07:03:19,339 - mmdet - INFO - Evaluating bbox...
2020-08-12 07:04:07,417 - mmdet - INFO - Epoch [2][181/181]	lr: 0.02444, bbox_mAP: 0.1600, bbox_mAP_50: 0.2810, bbox_mAP_75: 0.1490, bbox_mAP_s: 0.0040, bbox_mAP_m: 0.0440, bbox_mAP_l: 0.2000, bbox_mAP_copypaste: 0.160 0.281 0.149 0.004 0.044 0.200
2020-08-12 07:08:54,081 - mmdet - INFO - Epoch [3][50/181]	lr: 0.02644, eta: 16:33:00, time: 5.732, data_time: 2.681, memory: 9577, loss_rpn_cls: 0.0888, loss_rpn_bbox: 0.0398, s0.loss_cls: 0.1057, s0.acc: 97.1361, s0.loss_bbox: 0.0664, s1.loss_cls: 0.0503, s1.acc: 97.1821, s1.loss_bbox: 0.0580, s2.loss_cls: 0.0246, s2.acc: 97.1685, s2.loss_bbox: 0.0345, loss: 0.4682
2020-08-12 07:13:38,865 - mmdet - INFO - Epoch [3][100/181]	lr: 0.02844, eta: 16:47:17, time: 5.694, data_time: 2.553, memory: 9577, loss_rpn_cls: 0.0732, loss_rpn_bbox: 0.0394, s0.loss_cls: 0.1064, s0.acc: 97.1447, s0.loss_bbox: 0.0640, s1.loss_cls: 0.0514, s1.acc: 97.1648, s1.loss_bbox: 0.0584, s2.loss_cls: 0.0251, s2.acc: 97.2026, s2.loss_bbox: 0.0344, loss: 0.4523
2020-08-12 07:18:36,678 - mmdet - INFO - Epoch [3][150/181]	lr: 0.03000, eta: 17:03:04, time: 5.956, data_time: 2.860, memory: 9577, loss_rpn_cls: 0.0694, loss_rpn_bbox: 0.0352, s0.loss_cls: 0.1058, s0.acc: 96.9936, s0.loss_bbox: 0.0677, s1.loss_cls: 0.0525, s1.acc: 97.0047, s1.loss_bbox: 0.0636, s2.loss_cls: 0.0255, s2.acc: 97.0066, s2.loss_bbox: 0.0380, loss: 0.4577
2020-08-12 07:35:27,624 - mmdet - INFO - Evaluating bbox...
2020-08-12 07:36:22,429 - mmdet - INFO - Epoch [3][181/181]	lr: 0.03000, bbox_mAP: 0.1870, bbox_mAP_50: 0.3380, bbox_mAP_75: 0.1690, bbox_mAP_s: 0.0120, bbox_mAP_m: 0.0630, bbox_mAP_l: 0.2240, bbox_mAP_copypaste: 0.187 0.338 0.169 0.012 0.063 0.224
2020-08-12 07:41:09,365 - mmdet - INFO - Epoch [4][50/181]	lr: 0.03000, eta: 16:14:49, time: 5.738, data_time: 2.586, memory: 9577, loss_rpn_cls: 0.0758, loss_rpn_bbox: 0.0384, s0.loss_cls: 0.1012, s0.acc: 97.1593, s0.loss_bbox: 0.0677, s1.loss_cls: 0.0487, s1.acc: 97.2184, s1.loss_bbox: 0.0618, s2.loss_cls: 0.0237, s2.acc: 97.2801, s2.loss_bbox: 0.0375, loss: 0.4549
2020-08-12 07:45:55,385 - mmdet - INFO - Epoch [4][100/181]	lr: 0.03000, eta: 16:24:26, time: 5.718, data_time: 2.568, memory: 9577, loss_rpn_cls: 0.0629, loss_rpn_bbox: 0.0354, s0.loss_cls: 0.0953, s0.acc: 97.1929, s0.loss_bbox: 0.0665, s1.loss_cls: 0.0469, s1.acc: 97.2090, s1.loss_bbox: 0.0633, s2.loss_cls: 0.0234, s2.acc: 97.1712, s2.loss_bbox: 0.0400, loss: 0.4336
2020-08-12 07:50:49,482 - mmdet - INFO - Epoch [4][150/181]	lr: 0.03000, eta: 16:34:19, time: 5.882, data_time: 2.645, memory: 9577, loss_rpn_cls: 0.0824, loss_rpn_bbox: 0.0427, s0.loss_cls: 0.1065, s0.acc: 96.9019, s0.loss_bbox: 0.0688, s1.loss_cls: 0.0525, s1.acc: 96.9026, s1.loss_bbox: 0.0668, s2.loss_cls: 0.0259, s2.acc: 96.8715, s2.loss_bbox: 0.0404, loss: 0.4860
2020-08-12 08:07:37,590 - mmdet - INFO - Evaluating bbox...
2020-08-12 08:08:38,401 - mmdet - INFO - Epoch [4][181/181]	lr: 0.03000, bbox_mAP: 0.2090, bbox_mAP_50: 0.3710, bbox_mAP_75: 0.1900, bbox_mAP_s: 0.0100, bbox_mAP_m: 0.0690, bbox_mAP_l: 0.2610, bbox_mAP_copypaste: 0.209 0.371 0.190 0.010 0.069 0.261
2020-08-12 08:13:24,080 - mmdet - INFO - Epoch [5][50/181]	lr: 0.03000, eta: 15:57:25, time: 5.713, data_time: 2.741, memory: 9577, loss_rpn_cls: 0.0593, loss_rpn_bbox: 0.0369, s0.loss_cls: 0.0926, s0.acc: 97.2463, s0.loss_bbox: 0.0648, s1.loss_cls: 0.0454, s1.acc: 97.2340, s1.loss_bbox: 0.0646, s2.loss_cls: 0.0226, s2.acc: 97.1267, s2.loss_bbox: 0.0406, loss: 0.4269
2020-08-12 08:18:04,086 - mmdet - INFO - Epoch [5][100/181]	lr: 0.03000, eta: 16:02:37, time: 5.600, data_time: 2.508, memory: 9577, loss_rpn_cls: 0.0622, loss_rpn_bbox: 0.0332, s0.loss_cls: 0.0879, s0.acc: 97.4800, s0.loss_bbox: 0.0599, s1.loss_cls: 0.0428, s1.acc: 97.5127, s1.loss_bbox: 0.0580, s2.loss_cls: 0.0213, s2.acc: 97.4200, s2.loss_bbox: 0.0369, loss: 0.4023
2020-08-12 08:22:51,868 - mmdet - INFO - Epoch [5][150/181]	lr: 0.03000, eta: 16:08:26, time: 5.754, data_time: 2.668, memory: 9577, loss_rpn_cls: 0.0710, loss_rpn_bbox: 0.0383, s0.loss_cls: 0.0988, s0.acc: 97.0189, s0.loss_bbox: 0.0703, s1.loss_cls: 0.0475, s1.acc: 97.0927, s1.loss_bbox: 0.0670, s2.loss_cls: 0.0232, s2.acc: 97.1172, s2.loss_bbox: 0.0420, loss: 0.4580
2020-08-12 08:39:40,883 - mmdet - INFO - Evaluating bbox...
2020-08-12 08:40:31,163 - mmdet - INFO - Epoch [5][181/181]	lr: 0.03000, bbox_mAP: 0.2360, bbox_mAP_50: 0.4150, bbox_mAP_75: 0.2170, bbox_mAP_s: 0.0070, bbox_mAP_m: 0.0860, bbox_mAP_l: 0.2960, bbox_mAP_copypaste: 0.236 0.415 0.217 0.007 0.086 0.296
2020-08-12 08:45:21,368 - mmdet - INFO - Epoch [6][50/181]	lr: 0.03000, eta: 15:39:32, time: 5.803, data_time: 2.666, memory: 9577, loss_rpn_cls: 0.0604, loss_rpn_bbox: 0.0338, s0.loss_cls: 0.0846, s0.acc: 97.4355, s0.loss_bbox: 0.0604, s1.loss_cls: 0.0406, s1.acc: 97.5277, s1.loss_bbox: 0.0619, s2.loss_cls: 0.0205, s2.acc: 97.4696, s2.loss_bbox: 0.0405, loss: 0.4027
2020-08-12 08:50:07,877 - mmdet - INFO - Epoch [6][100/181]	lr: 0.03000, eta: 15:44:23, time: 5.728, data_time: 2.526, memory: 9577, loss_rpn_cls: 0.0584, loss_rpn_bbox: 0.0338, s0.loss_cls: 0.0895, s0.acc: 97.1883, s0.loss_bbox: 0.0685, s1.loss_cls: 0.0437, s1.acc: 97.2163, s1.loss_bbox: 0.0674, s2.loss_cls: 0.0221, s2.acc: 97.1512, s2.loss_bbox: 0.0435, loss: 0.4269
2020-08-12 08:55:00,874 - mmdet - INFO - Epoch [6][150/181]	lr: 0.03000, eta: 15:49:33, time: 5.860, data_time: 2.739, memory: 9577, loss_rpn_cls: 0.0614, loss_rpn_bbox: 0.0345, s0.loss_cls: 0.0907, s0.acc: 97.2104, s0.loss_bbox: 0.0671, s1.loss_cls: 0.0441, s1.acc: 97.2585, s1.loss_bbox: 0.0666, s2.loss_cls: 0.0220, s2.acc: 97.1825, s2.loss_bbox: 0.0427, loss: 0.4291
2020-08-12 09:11:19,756 - mmdet - INFO - Evaluating bbox...
2020-08-12 09:12:11,096 - mmdet - INFO - Epoch [6][181/181]	lr: 0.03000, bbox_mAP: 0.2390, bbox_mAP_50: 0.4150, bbox_mAP_75: 0.2220, bbox_mAP_s: 0.0080, bbox_mAP_m: 0.0930, bbox_mAP_l: 0.2960, bbox_mAP_copypaste: 0.239 0.415 0.222 0.008 0.093 0.296
2020-08-12 09:16:57,177 - mmdet - INFO - Epoch [7][50/181]	lr: 0.03000, eta: 15:24:05, time: 5.721, data_time: 2.754, memory: 9577, loss_rpn_cls: 0.0546, loss_rpn_bbox: 0.0338, s0.loss_cls: 0.0834, s0.acc: 97.3792, s0.loss_bbox: 0.0662, s1.loss_cls: 0.0406, s1.acc: 97.4057, s1.loss_bbox: 0.0672, s2.loss_cls: 0.0210, s2.acc: 97.2666, s2.loss_bbox: 0.0446, loss: 0.4115
2020-08-12 09:21:39,889 - mmdet - INFO - Epoch [7][100/181]	lr: 0.03000, eta: 15:26:55, time: 5.654, data_time: 2.589, memory: 9577, loss_rpn_cls: 0.0559, loss_rpn_bbox: 0.0344, s0.loss_cls: 0.0848, s0.acc: 97.3949, s0.loss_bbox: 0.0633, s1.loss_cls: 0.0417, s1.acc: 97.4099, s1.loss_bbox: 0.0643, s2.loss_cls: 0.0208, s2.acc: 97.3486, s2.loss_bbox: 0.0413, loss: 0.4064
2020-08-12 09:26:31,179 - mmdet - INFO - Epoch [7][150/181]	lr: 0.03000, eta: 15:30:28, time: 5.826, data_time: 2.787, memory: 9577, loss_rpn_cls: 0.0519, loss_rpn_bbox: 0.0340, s0.loss_cls: 0.0852, s0.acc: 97.3091, s0.loss_bbox: 0.0665, s1.loss_cls: 0.0407, s1.acc: 97.3821, s1.loss_bbox: 0.0660, s2.loss_cls: 0.0205, s2.acc: 97.3383, s2.loss_bbox: 0.0434, loss: 0.4081
2020-08-12 09:43:10,185 - mmdet - INFO - Evaluating bbox...
2020-08-12 09:44:15,253 - mmdet - INFO - Epoch [7][181/181]	lr: 0.03000, bbox_mAP: 0.2450, bbox_mAP_50: 0.4320, bbox_mAP_75: 0.2310, bbox_mAP_s: 0.0100, bbox_mAP_m: 0.1040, bbox_mAP_l: 0.3070, bbox_mAP_copypaste: 0.245 0.432 0.231 0.010 0.104 0.307
2020-08-12 09:48:59,279 - mmdet - INFO - Epoch [8][50/181]	lr: 0.03000, eta: 15:07:50, time: 5.678, data_time: 2.674, memory: 9577, loss_rpn_cls: 0.0527, loss_rpn_bbox: 0.0322, s0.loss_cls: 0.0812, s0.acc: 97.4064, s0.loss_bbox: 0.0642, s1.loss_cls: 0.0387, s1.acc: 97.5179, s1.loss_bbox: 0.0655, s2.loss_cls: 0.0195, s2.acc: 97.4489, s2.loss_bbox: 0.0420, loss: 0.3960
2020-08-12 09:53:41,386 - mmdet - INFO - Epoch [8][100/181]	lr: 0.03000, eta: 15:09:40, time: 5.643, data_time: 2.634, memory: 9577, loss_rpn_cls: 0.0488, loss_rpn_bbox: 0.0314, s0.loss_cls: 0.0846, s0.acc: 97.3434, s0.loss_bbox: 0.0661, s1.loss_cls: 0.0412, s1.acc: 97.4464, s1.loss_bbox: 0.0667, s2.loss_cls: 0.0210, s2.acc: 97.2723, s2.loss_bbox: 0.0428, loss: 0.4026
2020-08-12 09:58:33,264 - mmdet - INFO - Epoch [8][150/181]	lr: 0.03000, eta: 15:12:18, time: 5.834, data_time: 2.821, memory: 9577, loss_rpn_cls: 0.0542, loss_rpn_bbox: 0.0333, s0.loss_cls: 0.0886, s0.acc: 97.2471, s0.loss_bbox: 0.0672, s1.loss_cls: 0.0427, s1.acc: 97.3618, s1.loss_bbox: 0.0678, s2.loss_cls: 0.0214, s2.acc: 97.3200, s2.loss_bbox: 0.0447, loss: 0.4198
2020-08-12 10:15:22,268 - mmdet - INFO - Evaluating bbox...
2020-08-12 10:16:13,876 - mmdet - INFO - Epoch [8][181/181]	lr: 0.03000, bbox_mAP: 0.2490, bbox_mAP_50: 0.4370, bbox_mAP_75: 0.2290, bbox_mAP_s: 0.0080, bbox_mAP_m: 0.1030, bbox_mAP_l: 0.3170, bbox_mAP_copypaste: 0.249 0.437 0.229 0.008 0.103 0.317
2020-08-12 10:21:02,874 - mmdet - INFO - Epoch [9][50/181]	lr: 0.03000, eta: 14:52:39, time: 5.778, data_time: 2.776, memory: 9577, loss_rpn_cls: 0.0504, loss_rpn_bbox: 0.0329, s0.loss_cls: 0.0854, s0.acc: 97.3101, s0.loss_bbox: 0.0676, s1.loss_cls: 0.0414, s1.acc: 97.4078, s1.loss_bbox: 0.0679, s2.loss_cls: 0.0207, s2.acc: 97.3167, s2.loss_bbox: 0.0431, loss: 0.4094
2020-08-12 10:25:48,757 - mmdet - INFO - Epoch [9][100/181]	lr: 0.03000, eta: 14:54:12, time: 5.718, data_time: 2.666, memory: 9577, loss_rpn_cls: 0.0474, loss_rpn_bbox: 0.0325, s0.loss_cls: 0.0762, s0.acc: 97.5338, s0.loss_bbox: 0.0633, s1.loss_cls: 0.0374, s1.acc: 97.5894, s1.loss_bbox: 0.0666, s2.loss_cls: 0.0195, s2.acc: 97.4188, s2.loss_bbox: 0.0439, loss: 0.3869
2020-08-12 10:30:46,152 - mmdet - INFO - Epoch [9][150/181]	lr: 0.03000, eta: 14:56:40, time: 5.948, data_time: 2.987, memory: 9577, loss_rpn_cls: 0.0506, loss_rpn_bbox: 0.0321, s0.loss_cls: 0.0855, s0.acc: 97.2961, s0.loss_bbox: 0.0669, s1.loss_cls: 0.0418, s1.acc: 97.3394, s1.loss_bbox: 0.0660, s2.loss_cls: 0.0212, s2.acc: 97.2348, s2.loss_bbox: 0.0428, loss: 0.4068
2020-08-12 10:47:19,115 - mmdet - INFO - Evaluating bbox...
2020-08-12 10:48:10,447 - mmdet - INFO - Epoch [9][181/181]	lr: 0.03000, bbox_mAP: 0.2740, bbox_mAP_50: 0.4670, bbox_mAP_75: 0.2620, bbox_mAP_s: 0.0140, bbox_mAP_m: 0.1040, bbox_mAP_l: 0.3560, bbox_mAP_copypaste: 0.274 0.467 0.262 0.014 0.104 0.356
2020-08-12 10:52:58,184 - mmdet - INFO - Epoch [10][50/181]	lr: 0.03000, eta: 14:38:33, time: 5.754, data_time: 2.675, memory: 9577, loss_rpn_cls: 0.0498, loss_rpn_bbox: 0.0316, s0.loss_cls: 0.0825, s0.acc: 97.3278, s0.loss_bbox: 0.0672, s1.loss_cls: 0.0399, s1.acc: 97.4513, s1.loss_bbox: 0.0686, s2.loss_cls: 0.0203, s2.acc: 97.3490, s2.loss_bbox: 0.0448, loss: 0.4048
2020-08-12 10:57:41,158 - mmdet - INFO - Epoch [10][100/181]	lr: 0.03000, eta: 14:39:06, time: 5.660, data_time: 2.543, memory: 9577, loss_rpn_cls: 0.0431, loss_rpn_bbox: 0.0272, s0.loss_cls: 0.0760, s0.acc: 97.5802, s0.loss_bbox: 0.0621, s1.loss_cls: 0.0372, s1.acc: 97.6028, s1.loss_bbox: 0.0662, s2.loss_cls: 0.0191, s2.acc: 97.4789, s2.loss_bbox: 0.0431, loss: 0.3739
2020-08-12 11:02:32,669 - mmdet - INFO - Epoch [10][150/181]	lr: 0.03000, eta: 14:40:14, time: 5.829, data_time: 2.749, memory: 9577, loss_rpn_cls: 0.0453, loss_rpn_bbox: 0.0296, s0.loss_cls: 0.0775, s0.acc: 97.4704, s0.loss_bbox: 0.0650, s1.loss_cls: 0.0371, s1.acc: 97.6160, s1.loss_bbox: 0.0673, s2.loss_cls: 0.0187, s2.acc: 97.5282, s2.loss_bbox: 0.0439, loss: 0.3844
2020-08-12 11:19:22,355 - mmdet - INFO - Evaluating bbox...
2020-08-12 11:20:15,121 - mmdet - INFO - Epoch [10][181/181]	lr: 0.03000, bbox_mAP: 0.2690, bbox_mAP_50: 0.4660, bbox_mAP_75: 0.2600, bbox_mAP_s: 0.0140, bbox_mAP_m: 0.0970, bbox_mAP_l: 0.3450, bbox_mAP_copypaste: 0.269 0.466 0.260 0.014 0.097 0.345
2020-08-12 11:25:02,954 - mmdet - INFO - Epoch [11][50/181]	lr: 0.03000, eta: 14:23:30, time: 5.754, data_time: 2.664, memory: 9577, loss_rpn_cls: 0.0393, loss_rpn_bbox: 0.0309, s0.loss_cls: 0.0786, s0.acc: 97.3519, s0.loss_bbox: 0.0678, s1.loss_cls: 0.0386, s1.acc: 97.3800, s1.loss_bbox: 0.0718, s2.loss_cls: 0.0199, s2.acc: 97.2387, s2.loss_bbox: 0.0492, loss: 0.3960
2020-08-12 11:29:47,153 - mmdet - INFO - Epoch [11][100/181]	lr: 0.03000, eta: 14:23:42, time: 5.685, data_time: 2.519, memory: 9577, loss_rpn_cls: 0.0453, loss_rpn_bbox: 0.0306, s0.loss_cls: 0.0761, s0.acc: 97.4800, s0.loss_bbox: 0.0621, s1.loss_cls: 0.0363, s1.acc: 97.6042, s1.loss_bbox: 0.0638, s2.loss_cls: 0.0189, s2.acc: 97.4638, s2.loss_bbox: 0.0416, loss: 0.3746
2020-08-12 11:34:40,877 - mmdet - INFO - Epoch [11][150/181]	lr: 0.03000, eta: 14:24:30, time: 5.876, data_time: 2.723, memory: 9577, loss_rpn_cls: 0.0422, loss_rpn_bbox: 0.0300, s0.loss_cls: 0.0759, s0.acc: 97.5438, s0.loss_bbox: 0.0632, s1.loss_cls: 0.0378, s1.acc: 97.5861, s1.loss_bbox: 0.0665, s2.loss_cls: 0.0195, s2.acc: 97.4316, s2.loss_bbox: 0.0438, loss: 0.3790
2020-08-12 11:51:23,133 - mmdet - INFO - Evaluating bbox...
2020-08-12 11:52:21,439 - mmdet - INFO - Epoch [11][181/181]	lr: 0.03000, bbox_mAP: 0.3020, bbox_mAP_50: 0.5070, bbox_mAP_75: 0.2920, bbox_mAP_s: 0.0140, bbox_mAP_m: 0.1270, bbox_mAP_l: 0.3840, bbox_mAP_copypaste: 0.302 0.507 0.292 0.014 0.127 0.384
2020-08-12 11:57:05,766 - mmdet - INFO - Epoch [12][50/181]	lr: 0.03000, eta: 14:08:35, time: 5.685, data_time: 2.752, memory: 9577, loss_rpn_cls: 0.0409, loss_rpn_bbox: 0.0271, s0.loss_cls: 0.0761, s0.acc: 97.4320, s0.loss_bbox: 0.0639, s1.loss_cls: 0.0369, s1.acc: 97.5314, s1.loss_bbox: 0.0673, s2.loss_cls: 0.0193, s2.acc: 97.3844, s2.loss_bbox: 0.0448, loss: 0.3764

What should I do to implement the segmenation result?

I wonder how to approach this project. I think the model already has the segmentation result, isn't it?

请问您是怎么计算出cas的三个head的iou？

您好！
您在readme中说到“根据比赛0.1 0.3 0.5的iou要求, 将cas三个head的iou阈值调整为0.4 0.5 0.6“
请问您是怎么通过（0.1， 0.3，0.5）算出cas的（0.4，0.5，0.6)的呢？能详细的讲解一下吗？
谢谢您！

In the training results, the recall is very high, and the precision is very low

Thanks to you, I have succeeded in training and performing the model.

I received through baidu, 17,000 annotations from the training data, and divided them into 9:1 and conducted training and testing.

Nothing has changed from the model, but the test results are not sure if it's normal or not.

Could you give me some advice?

My optimizer setting is only one GPU, so I set it as below.

optimizer = dict(type='SGD', lr=0.00125, momentum=0.9, weight_decay=0.0001)

对于正常图像（不包含缺陷）图像怎么处理的呢

您好，请问对于正常图像（不包含缺陷）图像怎么处理的呢？看了下代码好像直接给排除掉了？

求助训练到最后有no such file or directory错误

我在train.sh里的设置：

求教为什么我无法保存权重，非常感谢

inference/merger_json2.py 里面 json_acc_path 这个参数为定义

请问作者当时网络到达多少代数才收敛的？

关于测试

运行test.py时候报错 if not isinstance(outputs[0], dict):
IndexError: list index out of range
completed: 0, elapsed: 0s
似乎是找不到图片，所以outputs列表中没有任何元素，我试过很多图片路径的改变都不行

Not perform as expect

Hi，After I trained the models follow your tips , I found that this two models not perform very well , even I do the test on a few train set data。I suspect that something is wrong in my training？ Because I noticed that my loss not drop sharply as I do in other train works。Can you give me some suggestion？ Thanks！
Best wishes！

你好，请问一下在测试的时候缺少json文件

缺少以下文件
result_map.bbox.json
result_zj.bbox.json

"CalledProcessError" when training data on my workstation

I use a new machine(CentOs7) and the program return CalledProcessError when running dist_train.sh:

this is train.sh file:

and dist_train.sh:

All dependencies install succeed. cudnn version 7.6.3

Could you please help me figure out the problem? Thank you!

an issue

Hello.
I'm very interested in your research.
But I only have win10 system here.
Is the system suitable for your research?

数据集中部分标注似乎有问题

数据集中有三种高度的图像，分别是 1696、1800和1810.
对于高度为1800和1810的，标注框在图像底部时，会被截断为1696.
存在超过1696的情况是因为标注框 [x1, y1, x2, y2] 中的y1没有被截断，只有y2被截断。
大家都是直接用这份数据的标注吗？还是需要进行进行一些调整？

复赛数据集

你好，百度网盘里只有4000多张图片，但我看官网说花色布数据约12000张，请问你有全的复赛数据吗？如果有，可否分享一下，谢谢了

train error

I only have 1gpu
when I run CUDA_VISIBLE_DEVICES=0 ./dist_train.sh ../config/fabric_defect/cascade_rcnn_r50_fpn_70e.py 1

how can I fix it?

training error,Help!

When I run train.sh the following error occurs

Especially this sentence：RuntimeError: While copying the parameter named bbox_head.0.fc_cls.weight, whose dimensions in the model are torch.Size([16, 1024]) and whose dimensions in the checkpoint are torch.Size([81, 1024]).
Point out that the error is that the dimensions do not match each other, but the weights are downloaded according to the URL in train.sh ,How can I solve this error?

可以提供一下初赛的数据集吗？

您好！请问可以提供一下初赛的数据集吗？谢谢！

Leverage in datasets with different backgrounds

question about using the model. It seems that the model uses a template image to detect the difference by comparing the differences. Then, I don't think model can respond to the situation where the background changes every time like DAGM dataset, am I approaching it wrong?

使用mmdetection官方开源的casacde-rcnn-r50-fpn-2x的COCO预训练模型
下载预训练模型后进行转换变为支持CFRCNN模型的预训练模型`
coco的预训练模型是81类别的，而本数据是16类的，我看到对checkpoint文件转换也没有处理这个问题，那么预训练模型是怎么能成功加载的？
谢谢！

epoch个数设置

请问在该比赛中，epcoh设置为12和70效果相差的大吗？

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.