search

tianweiy / CenterPoint

MIT License
1.85k stars 450 forks source link

problem on the pub_test.py #368

Open JimmyGea opened 1 year ago

JimmyGea commented 1 year ago

Hi,thank you for your great work. l try to follow your work on Getting Started with CenterPoint on nuScenes,when l running this command: python tools/nusc_tracking/pub_test.py --work_dir work_dirs/nusc_centerpoint_voxelnet_0075voxel_fix_bn_z/output/outputs --checkpoint work_dirs/nusc_centerpoint_voxelnet_0075voxel_fix_bn_z/output/infos_val_10sweeps_withvelo_filter_True.json The following error occurred during the pub_test.

Traceback (most recent call last): File "tools/nusc_tracking/pub_test.py", line 194, in eval_tracking() File "tools/nusc_tracking/pub_test.py", line 161, in eval_tracking args.root File "tools/nusc_tracking/pub_test.py", line 179, in eval metrics_summary = nus_eval.main() File "/home/xuejunming/CenterPoint/nuscenes-devkit/python-sdk/nuscenes/eval/tracking/evaluate.py", line 205, in main metrics, metric_data_list = self.evaluate() File "/home/xuejunming/CenterPoint/nuscenes-devkit/python-sdk/nuscenes/eval/tracking/evaluate.py", line 135, in evaluate accumulate_class(class_name) File "/home/xuejunming/CenterPoint/nuscenes-devkit/python-sdk/nuscenes/eval/tracking/evaluate.py", line 131, in accumulate_class curr_md = curr_ev.accumulate() File "/home/xuejunming/CenterPoint/nuscenes-devkit/python-sdk/nuscenes/eval/tracking/algo.py", line 156, in accumulate assert unachieved_thresholds + duplicate_thresholds + len(thresh_metrics) == self.num_thresholds AssertionError

error

Thanks for your kind attention and look forward your prompt reply.

JimmyGea commented 1 year ago

If I may ask one more question, how to create infos_train_10sweeps_withvelo_filter_True.json? I can only create infos_val_10sweeps_withvelo_filter_True.json according to your instructions.

tianweiy commented 1 year ago

for the first question, maybe related to https://github.com/tianweiy/CenterPoint/issues/292#issuecomment-1105121604 ?

tianweiy commented 1 year ago

to generate detection on training set, modify val anno path to info_train https://github.com/tianweiy/CenterPoint/blob/d3a248fa56db2601860d576d5934d00fee9916eb/configs/nusc/voxelnet/nusc_centerpoint_voxelnet_0075voxel_fix_bn_z_flip.py#L171

tianweiy commented 1 year ago

you can also directly use prediction from https://github.com/tianweiy/CenterPoint/issues/249

JimmyGea commented 1 year ago

to generate detection on training set, modify val anno path to info_train

https://github.com/tianweiy/CenterPoint/blob/d3a248fa56db2601860d576d5934d00fee9916eb/configs/nusc/voxelnet/nusc_centerpoint_voxelnet_0075voxel_fix_bn_z_flip.py#L171

Thank you!