nuscenes data visualization fails

[Ultran-cyf]

Can anyone tell me how to fix it?

command:
python tools/demo.py --cfg_file tools/cfgs/dsvt_models/dsvt_plain_1f_onestage_nusences.yaml --data_path demo --ckpt ckpt/DSVT_Nuscenes_val.pth

The following error occurs when I run demo.py to test the Lidar data for nuscenes:

2024-07-31 10:22:55,076   INFO  -----------------Quick Demo of OpenPCDet-------------------------
2024-07-31 10:22:55,076   INFO  Total number of samples:        3
/home/user/anaconda3/envs/dsvt/lib/python3.9/site-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1695392035629/work/aten/src/ATen/native/TensorShape.cpp:3526.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
2024-07-31 10:22:55,186   INFO  ==> Loading parameters from checkpoint ckpt/DSVT_Nuscenes_val.pth to CPU
2024-07-31 10:22:55,213   INFO  ==> Done (loaded 449/449)
Traceback (most recent call last):
  File "/home/user/Pycharm_Project/DSVT-master/tools/demo.py", line 112, in <module>
    main()
  File "/home/user/Pycharm_Project/DSVT-master/tools/demo.py", line 94, in main
    for idx, data_dict in enumerate(demo_dataset):
  File "/home/user/Pycharm_Project/DSVT-master/tools/demo.py", line 59, in __getitem__
    data_dict = self.prepare_data(data_dict=input_dict)
  File "/home/user/Pycharm_Project/DSVT-master/pcdet/datasets/dataset.py", line 180, in prepare_data
    data_dict = self.point_feature_encoder.forward(data_dict)
  File "/home/user/Pycharm_Project/DSVT-master/pcdet/datasets/processor/point_feature_encoder.py", line 29, in forward
    data_dict['points'], use_lead_xyz = getattr(self, self.point_encoding_config.encoding_type)(
  File "/home/user/Pycharm_Project/DSVT-master/pcdet/datasets/processor/point_feature_encoder.py", line 48, in absolute_coordinates_encoding
    assert points.shape[-1] == len(self.src_feature_list)
AssertionError

I've tried removing the 'timestamp' field from POINT_FEATURE_ENCODING in nuscenes_dataset.yaml, but it still doesn't run!

POINT_FEATURE_ENCODING: { encoding_type: absolute_coordinates_encoding, used_feature_list: ['x', 'y', 'z', 'intensity', 'timestamp'], src_feature_list: ['x', 'y', 'z', 'intensity', 'timestamp'], }

The error is as follows:

2024-07-31 10:34:15,288   INFO  -----------------Quick Demo of OpenPCDet-------------------------
2024-07-31 10:34:15,288   INFO  Total number of samples: 	3
/home/user/anaconda3/envs/dsvt/lib/python3.9/site-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1695392035629/work/aten/src/ATen/native/TensorShape.cpp:3526.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
2024-07-31 10:34:15,368   INFO  ==> Loading parameters from checkpoint /home/user/Pycharm_Project/DSVT-master/ckpt/DSVT_Nuscenes_val.pth to CPU
2024-07-31 10:34:15,387   INFO  Not updated weight vfe.pfn_layers.0.linear.weight: torch.Size([64, 10])
2024-07-31 10:34:15,387   INFO  ==> Done (loaded 448/449)
2024-07-31 10:34:15,543   INFO  Visualized sample index: 	1
Traceback (most recent call last):
  File "/home/user/Pycharm_Project/DSVT-master/tools/demo.py", line 112, in <module>
    main()
  File "/home/user/Pycharm_Project/DSVT-master/tools/demo.py", line 100, in main
    V.draw_scenes(
  File "/home/user/Pycharm_Project/DSVT-master/tools/visual_utils/open3d_vis_utils.py", line 70, in draw_scenes
    vis = draw_box(vis, ref_boxes, (0, 1, 0), ref_labels, ref_scores)
  File "/home/user/Pycharm_Project/DSVT-master/tools/visual_utils/open3d_vis_utils.py", line 109, in draw_box
    line_set.paint_uniform_color(box_colormap[ref_labels[i]])
IndexError: list index out of range

[chenshi3]

Please note that the demo.py file is cloned directly from the original pcdet repository and has not been validated or tested by our team.

[wwwyyyddd]

@Ultran-cyf The out of range error occurs from the global "box_colormap" def in "open3d_vis_utils.py", which is less than box target categories of Nuscences, you could easily mod them yourself to get expected visual.