Sparse4D有三个版本其中V1和V2版本的官方文档中环境依赖写得比较模糊且依赖库有版本冲突。1. Sparse4D V1创建环境conda create sparse4dv1 python3.8激活环境conda activate sparse4dv1安装torch, torchvision, torchaudiopip install torch1.13.0cu116 torchvision0.14.0cu116 torchaudio0.13.0 --extra-index-url https://download.pytorch.org/whl/cu116安装mmcvpip install mmcv_full1.7.1安装mmdet3dpip install spconv-cu1162.2.4 pip install mmdet3d1.0.0rc1安装其它依赖pip install mmdet2.28.2 urllib31.26.16 pyquaternion0.9.9 nuscenes-devkit1.1.10 yapf0.33.0 tensorboard2.14.0 motmetrics1.1.3 pandas1.1.52. Sparse4D V2创建环境conda create sparse4dv2 python3.8激活环境conda activate sparse4dv2安装torch, torchvision, torchaudiopip install torch1.13.0cu116 torchvision0.14.0cu116 torchaudio0.13.0 --extra-index-url https://download.pytorch.org/whl/cu116安装mmcvpip install mmcv_full1.7.1安装其它依赖pip install mmdet2.28.2 urllib31.26.16 pyquaternion0.9.9 nuscenes-devkit1.1.10 yapf0.33.0 tensorboard2.14.0 motmetrics1.1.3 pandas1.1.53. Sparse4D V33.1. 环境配置创建环境conda create sparse4dv3 python3.8激活环境conda activate sparse4dv3安装torch, torchvision, torchaudiopip install torch1.13.0cu116 torchvision0.14.0cu116 torchaudio0.13.0 --extra-index-url https://download.pytorch.org/whl/cu116安装mmcvpip install mmcv_full1.7.1安装mmdet3d_plugincd projects/mmdet3d_plugin/ops python3 setup.py develop cd ../../../安装其它依赖pip install mmdet2.28.2 urllib31.26.16 pyquaternion0.9.9 nuscenes-devkit1.1.10 yapf0.33.0 tensorboard2.14.0 motmetrics1.1.3 pandas1.1.53.2. 数据准备首先下载并解压nuscenes数据然后把它链接到本地cd ${sparse4d_path} mkdir data ln -s path/to/nuscenes ./data/nuscenes数据转换pkl_pathdata/nuscenes_anno_pkls mkdir -p ${pkl_path} python3 tools/nuscenes_converter.py --version v1.0-mini --info_prefix ${pkl_path}/nuscenes-mini python3 tools/nuscenes_converter.py --version v1.0-trainval,v1.0-test --info_prefix ${pkl_path}/nuscenes使用K-means生成锚点python3 tools/anchor_generator.py --ann_file ${pkl_path}/nuscenes_infos_train.pkl3.3. 训练与测试下载预训练的backbonemkdir ckpt wget https://download.pytorch.org/models/resnet50-19c8e357.pth -O ckpt/resnet50-19c8e357.pth训练与测试# train bash local_train.sh sparse4dv3_temporal_r50_1x8_bs6_256x704 # test bash local_test.sh sparse4dv3_temporal_r50_1x8_bs6_256x704 path/to/checkpoint参考文献GitHub - HorizonRobotics/Sparse4D
