1. Train camera-based baseline with 8 GPUs.
bash run.sh ./projects/baselines/CAM-R50_img1600_128x128x10.py 8
2. Train LiDAR-based baseline with 8 GPUs.
bash run.sh ./projects/baselines/LiDAR_128x128x10.py 8
3. Train multimodal baseline with 8 GPUs.
bash run.sh ./projects/baselines/Multimodal-R50_img1600_128x128x10.py 8
4. Train camera-based CONet with 8 GPUs.
bash run.sh ./projects/Cascade-Occupancy-Network/CAM-R50_img1600_cascade_x4.py 8
5. Train LiDAR-based CONet with 8 GPUs.
bash run.sh ./projects/Cascade-Occupancy-Network/LiDAR_cascade_x4.py 8
6. Train multimodal CONet with 8 GPUs.
bash run.sh ./projects/Cascade-Occupancy-Network/Multimodal-R50_img1600_cascade_x4.py 8
Evaluation example.
bash run_eval.sh $PATH_TO_CFG $PATH_TO_CKPT $GPU_NUM
Temporarily only support saving occupancy predictions (refer to MonoScene for visualization tools)
bash run_eval.sh $PATH_TO_CFG $PATH_TO_CKPT $GPU_NUM --show --show-dir $PATH