README.md

Collaborative Perceiver: Elevating Vision-based 3D Object Detection by Local Density-Aware Spatial Occupancy

Official implementation of Collaborative Perceiver: Elevating Vision-based 3D Object Detection by Local Density-Aware Spatial Occupancy.

Get started

Environment Installation

Please see install.md.

Data Preparation

Please see data_preparation.md.

The final total data structure is shown like this:

cop
├── mmdet3d_plugin
├── tools
├── configs
├── data
│   ├── nuscenes
│   │   ├── maps
│   │   ├── panoptic
│   │   ├── processdPoints
│   │   ├── ldo_occ
│   │   ├── samples
│   │   ├── sweeps
│   │   ├── v1.0-test
|   |   ├── v1.0-trainval
│   │   ├── nuscenes_infos_train.pkl
│   │   ├── nuscenes_infos_val.pkl
│   │   ├── nuscenes_infos_train_cop.pkl
│   │   ├── nuscenes_infos_val_cop.pkl

Training

For training process, we use config file in $cop/configs/cop to define model, dataset and hyber parameters. Run the following command to start a training process. For example:

tools/dist_train.sh configs/cop/cop_r101_cbgs.py 4 

Resume the training

tools/dist_test.sh configs/cop/cop_r50_cbgs.py 
                   --resume-from cop_r101_cbgs.pth 4  

Evaluation

For testing bbox scores, run the following command:

tools/dist_test.sh configs/cop/cop_r50_cbgs.py 
                   cop_r101_cbgs.pth 4 
                   --eval bbox  

Acknowledgements

Many thanks to the authors of open-mmlab, CenterPoint, Lift-Splat-Shoot, BEVDet and BEVDepth.

Citation

@article{yuan2025collaborativeperceiverelevatingvisionbased,
  title={Collaborative Perceiver: Elevating Vision-based 3D Object Detection via Local Density-Aware Spatial Occupancy},
  author={Jicheng Yuan and Manh Nguyen Duc and Qian Liu and Manfred Hauswirth and Danh Le Phuoc},
  journal={arXiv preprint arXiv:2507.21358},
  year={2025}
}