/Fast-BEV

Fast-BEV: Towards Real-time On-vehicle Bird’s-Eye View Perception

Primary LanguagePythonOtherNOASSERTION

Fast-BEV

Fast-BEV: A Fast and Strong Bird’s-Eye View Perception Baseline image image image

Usage

Installation

  • CUDA>=9.2, GCC>=5.4, Python >= 3.6, Pytorch >= 1.8.1, Torchvision >= 0.9.1

  • MMCV-full == 1.4.0, MMDetection == 2.14.0, MMSegmentation == 0.14.1

    # gcc >= 5.4 

cd env/mmcv
# TODO
MMCV_OPS=1 pip install -v . --user

cd ../mmdetection
pip install -v -e . --user

cd ../mmsegmentation
pip install -v -e . --user

cd ../../
pip install -v -e . --user

  • Other requirements

    pip install -r requirements.txt --user

Dataset preparation

Please download nuscenes dataset and organize them as follows:

TODO

If you are using ceph, you can change the arguments in the configuration.

e.g.

file_client_args = dict(
    backend='petrel',
    path_mapping=dict({
        data_root: 'ceph:s3://path/to/data'}))

train_pipeline = [
    dict(
        type='MultiViewPipeline', 
        sequential=True,
        n_images=6,
        n_times=4,
        transforms=[
            dict(
                type='LoadImageFromFile',
                file_client_args=file_client_args),
        ]),
    ...
]

Training

We provide several configs in configs/fastbev/exp/paper.

Configure the tools/fastbev_run.sh script like

slurm_train $PARTITION 32 paper/<CONFIG_NAME>

And run

sh tools/fastbev_run.sh <PARTITION>

Evaluation

  • Inference

    Configure the tools/fastbev_run.sh script like

    slurm_test $PARTITION 16 paper/<CONFIG_NAME>

    And run

    sh tools/fastbev_run.sh <PARTITION>

  • Evaluation

    Configure the tools/fastbev_run.sh script like

    slurm_eval $PARTITION 1 paper/<CONFIG_NAME>

    And run

    sh tools/fastbev_run.sh <PARTITION>

Deployment

TODO

View Transformation Latency on device

2D-to-3D on CUDA & CPU

Citation

@article{li2023fast,
  title={Fast-BEV: A Fast and Strong Bird's-Eye View Perception Baseline},
  author={Li, Yangguang and Huang, Bin and Chen, Zeren and Cui, Yufeng and Liang, Feng and Shen, Mingzhu and Liu, Fenggang and Xie, Enze and Sheng, Lu and Ouyang, Wanli and others},
  journal={arXiv preprint arXiv:2301.12511},
  year={2023}
}