Comments (17)
glad to hear. Pillar is our fastest / most inaccurate model so you can try other stronger model like https://github.com/tianweiy/CenterPoint/blob/master/configs/centerpoint/nusc_centerpoint_voxelnet_01voxel.py You also want to tune this function to get desired number of output
from centerpoint.
not sure about rivz details. Tracking is just detection + ID so I guess it is possible.
from centerpoint.
CenterPoint/tools/multi_sweep_inference.py
Line 187 in caa2f76
Change to shape = [grid_size]
from centerpoint.
@tianweiy Yes, I have already tried that, but still getting this error
[ERROR] [1595862420.362335]: bad callback: <function rslidar_callback at 0x7f56679602f0>
Traceback (most recent call last):
File "/opt/ros/kinetic/lib/python2.7/dist-packages/rospy/topics.py", line 750, in _invoke_callback
cb(msg)
File "/home/siddhantsahu/centerpoint_p3_ws/src/centerpoint_ros_node/src/tools/multi_sweep_inference.py", line 339, in rslidar_callback
scores, dt_box_lidar, types = proc_1.run()
File "/home/siddhantsahu/centerpoint_p3_ws/src/centerpoint_ros_node/src/tools/multi_sweep_inference.py", line 192, in run
outputs = self.net(self.inputs)[0] #this is where the error is
File "/home/siddhantsahu/miniconda3/envs/centerpoint/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in __call__
result = self.forward(*input, **kwargs)
File "/home/siddhantsahu/Desktop/CenterPoint/det3d/models/detectors/point_pillars.py", line 52, in forward
return self.bbox_head.loss(example, preds)
File "/home/siddhantsahu/Desktop/CenterPoint/det3d/models/bbox_heads/mg_head.py", line 1294, in loss
hm_loss = self.crit(preds_dict['hm'], example['hm'][task_id])
KeyError: 'hm'
from centerpoint.
change
CenterPoint/tools/multi_sweep_inference.py
Line 192 in caa2f76
to
outputs = self.net(self.inputs, return_loss=False)[0]
from centerpoint.
Still
[ERROR] [1595862876.321550]: bad callback: <function rslidar_callback at 0x7f56497ac2f0>
Traceback (most recent call last):
File "/opt/ros/kinetic/lib/python2.7/dist-packages/rospy/topics.py", line 750, in _invoke_callback
cb(msg)
File "/home/siddhantsahu/centerpoint_p3_ws/src/centerpoint_ros_node/src/tools/multi_sweep_inference.py", line 339, in rslidar_callback
scores, dt_box_lidar, types = proc_1.run()
File "/home/siddhantsahu/centerpoint_p3_ws/src/centerpoint_ros_node/src/tools/multi_sweep_inference.py", line 192, in run
outputs = self.net(self.inputs, return_loss=False)[0] #this is where the error is
File "/home/siddhantsahu/miniconda3/envs/centerpoint/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in __call__
result = self.forward(*input, **kwargs)
File "/home/siddhantsahu/Desktop/CenterPoint/det3d/models/detectors/point_pillars.py", line 54, in forward
return self.bbox_head.predict(example, preds, self.test_cfg)
File "/home/siddhantsahu/Desktop/CenterPoint/det3d/models/bbox_heads/mg_head.py", line 1459, in predict
task_id=task_id
File "/home/siddhantsahu/Desktop/CenterPoint/det3d/core/utils/center_utils.py", line 392, in ddd_decode
keep = _circle_nms(boxes, min_radius=cfg.min_radius[task_id], post_max_size=cfg.post_max_size)
File "/home/siddhantsahu/Desktop/CenterPoint/det3d/core/utils/center_utils.py", line 274, in _circle_nms
keep = np.array(circle_nms(boxes.cpu().numpy(), thresh=min_radius))[:post_max_size]
RuntimeError: Can't call numpy() on Variable that requires grad. Use var.detach().numpy() instead.
from centerpoint.
with torch.no_grad():
outputs = self.net(self.inputs, return_loss=False)[0]
from centerpoint.
Are there still any other errors?
from centerpoint.
@tianweiy Thanks a lot! Yes, there are no errors now and the predictions also seems to match with the annotation ground truth. But there appears to be more number of unnecessary predicted boxes.
The transparent boxes are ground truth and solid ones are predicted. I am using the centerpoint_pillar_512 trained model.
from centerpoint.
Sure, will try with the best trained models. Fine tuning the thresh
definitely helps.
The model which I tried nusc_centerpoint_pp_02voxel_circle_nms_demo is it only detection or detection + tracking as well? Can I try to use the centerpoint_voxel_1440_dcn_flip(Rotated NMS) tracking model and other tracking pre-trained models well?
from centerpoint.
only detection. Tracking script is here. You can use all pre-trained models but currently, the two ros scripts don't support the flip data augmentation.
from centerpoint.
In addition to detection can tracking be visualized in ROS rivz as well?
from centerpoint.
check the id info in output tensor and add it in visual box.
from centerpoint.
@muzi2045 Thanks for suggesting. Should a new ROS topic needs to created and the id data needs to be published there or it can be published to an existing ROS topic? And how to visualize the respective id over their bounding boxes?
from centerpoint.
check the id info in output tensor and add it in visual box.
The output tensor contains box3d_lidar, scores and label_preds. There is no id info. Can you provide more information, please?
from centerpoint.
@YoushaaMurhij Yeah, no id info in the tensor. You need to follow the tools/tracking/pub_test.py
to also generate id info in the ros inference file.
from centerpoint.
@tianweiy Thanks a lot! Yes, there are no errors now and the predictions also seems to match with the annotation ground truth. But there appears to be more number of unnecessary predicted boxes.
The transparent boxes are ground truth and solid ones are predicted. I am using the centerpoint_pillar_512 trained model.
HI @iamsiddhantsahu, one queston: I tried to reproduce the visualization of this screenshot in rviz, but only the predicted bounding boxes (i.e. only solid boxes) are shown, which come from the / pp_boxes topic. How did you manage to display also the ground truth boxes (transparent ones) at the same time? I didn't see any other topic from the rosbag file publishing this information.
I would appreciate your help, thanks.
from centerpoint.
Related Issues (20)
- Virtual Points Loading
- the error when i train nuscenes, i think the reason is my spconv>2.0, but how can i change the code to solve it?
- about centerpoint-pillar nuscenes test set mAP and NDS HOT 1
- Why do Waymo and NuScenes have different `us_layer_strides` even though they use the same CenterPoint?
- No such file or directory: 'data/nuScenes/infos_ Test_ 10 sweeps_ Withvelo_ Filter_ True. pkl
- Tracking bug with Hungarian
- RuntimeError: Error compiling objects for extension HOT 1
- Unable to find nuScenes devkit HOT 1
- requirements change
- Error when starting training: TypeError: string indices must be integers
- Unable to build iou3d_NMS
- When I run create_data.py use nuScenes dataset have something wrong
- What's the speed of tracker in WOD
- time
- visual pkl
- assert len(detections) == 6008
- 请问有用mini数据集得到的centerpoint的检测结果吗,类似这个infos_val_10sweeps_withvelo_filter_True.json文件
- Request for Combined Train and Validation Split File
- Personal dataset HOT 1
- Ground truth filtering
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.
from centerpoint.