Skip to content

Latest commit

 

History

History
31 lines (24 loc) · 1.56 KB

getting_started.md

File metadata and controls

31 lines (24 loc) · 1.56 KB

Getting started with DeepAccident

Training

To train the models on DeepAccident with 8 GPUs, run:

bash tools/dist_train.sh $CONFIG 8 --load-from ./data/DeepAccident_data/beverse_tiny.pth --work-dir $WORK_DIR
e.g. bash tools/dist_train.sh projects/configs/DeepAccident_tiny.py 8 --load-from ./data/DeepAccident_data/beverse_tiny.pth --work-dir work_dirs/DeepAccident_tiny

To train the models on DeepAccident with single GPU, run:

python tools/train.py $CONFIG --load-from ./data/DeepAccident_data/beverse_tiny.pth --work-dir $WORK_DIR
e.g. python tools/train.py projects/configs/DeepAccident_tiny.py --load-from ./data/DeepAccident_data/beverse_tiny.pth --work-dir work_dirs/DeepAccident_tiny

Evaluation

To evaluate the models on DeepAccident, run:

python tools/test.py $YOUR_CONFIG --checkpoint $YOUR_CKPT --eval=distance_mAP --mtl --tp-save_dir $TP_DIR_FOR_ACCIDENT_PREDICTION
e.g. python tools/test.py projects/configs/DeepAccident_tiny.py --checkpoint work_dirs/DeepAccident_tiny/latest_epoch.pth --eval=distance_mAP --mtl --tp-save-dir work_dirs/DeepAccident_tiny/accident_tp

Visualization

To visualize the predictions, run:

python tools/test.py $YOUR_CONFIG --checkpoint $YOUR_CKPT --eval=distance_mAP --mtl --tp-save_dir $TP_DIR_FOR_ACCIDENT_PREDICTION --show --show-dir $SHOW_DIR
e.g. python tools/test.py projects/configs/DeepAccident_tiny.py --checkpoint work_dirs/DeepAccident_tiny/latest_epoch.pth --eval=distance_mAP --mtl --tp-save-dir work_dirs/DeepAccident_tiny/accident_tp --show --show-dir work_dirs/DeepAccident_tiny