search

LittlePey / SFD

Sparse Fuse Dense: Towards High Quality 3D Detection with Depth Completion (CVPR 2022, Oral)
Apache License 2.0
263 stars 35 forks source link

KeyError: 'points_pseudo' when try the demo to eval and visualize the result. #28

Closed ChunZheng2022 closed 1 year ago

ChunZheng2022 commented 1 year ago

(cu113) lcz@wp:~/code/multi_modal/SFD/tools$ python demo.py --cfg_file cfgs/kitti_models/sfd.yaml --ckpt checkpoint_epoch_34.pth --data_path 000001.bin

Traceback (most recent call last): File "demo.py", line 103, in main() File "demo.py", line 87, in main for idx, data_dict in enumerate(demo_dataset): File "demo.py", line 52, in getitem data_dict = self.prepare_data(data_dict=input_dict) File "/home/lcz/code/multi_modal/SFD/pcdet/datasets/dataset.py", line 143, in prepare_data data_dict = self.point_feature_encoder.forward(data_dict) File "/home/lcz/code/multi_modal/SFD/pcdet/datasets/processor/point_feature_encoder.py", line 44, in forward data_dict['points_pseudo'] KeyError: 'points_pseudo'

The demo function in OpenPCDet currently does no support reading multimodal input yet.

swaggywilliam commented 10 months ago

I have met the same error, could you tell me how to solve it please? thank you!

ChunZheng2022 commented 10 months ago

@swaggywilliam Hi, the demo create a demo dataset for inference. And the default demo dataset does not support pseudo points. You can down addition vis tools like thishttps://github.com/hailanyi/3D-Detection-Tracking-Viewer to visualize ur results.

swaggywilliam commented 10 months ago

@swaggywilliam Hi, the demo create a demo dataset for inference. And the default demo dataset does not support pseudo points. You can down addition vis tools like thishttps://github.com/hailanyi/3D-Detection-Tracking-Viewer to visualize ur results.

您好，请问您的意思是sfd.yaml是使用伪雷达点云数据的，但是OpenPCDet的demo.py并不支持伪雷达点云数据，所以没法用以下命令?

python demo.py --cfg_file cfgs/kitti_models/sfd.yaml --ckpt checkpoint_epoch_34.pth --data_path 000001.bin
ChunZheng2022 commented 10 months ago

@swaggywilliam 是的，不过你可以使用这个推理脚本或者test.py来得到推理结果，并使用额外的工具进行可视化。

scripts/dist_test.sh 8 --cfg_file cfgs/kitti_models/sfd.yaml  --gpu_id 0,1,2,3,4,5,6,7 --batch_size 28 \
--ckpt ../output/kitti_models/sfd/default/ckpt/checkpoint_epoch_40.pth

通过这样的方法进行可视化需要添加--save_to_file命令。

ChunZheng2022 commented 10 months ago

可以参考这个 #13 @swaggywilliam

swaggywilliam commented 10 months ago

@swaggywilliam 是的，不过你可以使用这个推理脚本或者test.py来得到推理结果，并使用额外的工具进行可视化。

scripts/dist_test.sh 8 --cfg_file cfgs/kitti_models/sfd.yaml  --gpu_id 0,1,2,3,4,5,6,7 --batch_size 28 \
--ckpt ../output/kitti_models/sfd/default/ckpt/checkpoint_epoch_40.pth

通过这样的方法进行可视化需要添加--save_to_file命令。

非常感谢，那我先尝试把train.py跑起来