RCBEVDet项目复现

sudo docker pull hesoyam1377/rcbevdet_cuda116_final:v2

sudo docker run -it -d --name rcbevdet --gpus all --ipc=host -p 8888:8888 -v /home/added/projects/py/RCBEVDet/rcbevdet-master:/workspace/RCBEVDet -v /home/added/projects/dataset/nuScenes/v1.0-mini:/workspace/RCBEVDet/data/nuscenes 18c34ca45675


cd /workspace/RCBEVDet
TORCH_CUDA_ARCH_LIST="8.6" FORCE_CUDA=1 pip install -v -e . # 配置mmdet3D环境
python -c "import mmdet3d; print('MMDet3D Version:', mmdet3d.__version__)" # 测试mmdet3D环境

# 修改create_data_nuscenes_RC.py文件 1. mini 2. 去掉test

from mmdet3d.datasets import NuScenesDatasetRC as NuScenesDataset_R # 修改 tools/data_converter文件夹中的nuscenes_converter_RC.py文件第20行

python ./tools/create_data_nuscenes_RC.py # 生成文件nuscenes_RC_infos_train.pkl、nuscenes_RC_infos_val.pkl

bash tools/test.sh <config file> <checkpoint file> <gpu nums> --eval bbox --show --show-dir ./results

bash tools/dist_test.sh /root/rcbevdet-master/configs/rcbevdet/det-256x704-r50-BEV128-9kf-depth-cbgs12e-circlel
arger.py /root/rcbevdet-master/models/rcbevdet.pth 1 --eval bbox --show --show-dir ./results