- 4 NVIDIA 3090Ti GPUs (24GB memory each)
- 32 CPUs
pip install -r requirements.txt
Processed data and pretrained ResNet (50 and 152) can be downloaded from the links below. Put it in top-level folder.
Baidu Disk password: 8eqb
DFF, backbone ResNet-50
CASENet, backbone ResNet-152
CUDA_VISIBLE_DEVICES=0,1,2,3 python train.py --dataset simple --model DFF --backbone resnet50 --batch-size 8 --lr 0.001 --epochs 200 --crop-size 960 --kernel-size 5 --edge-weight 0.4
CUDA_VISIBLE_DEVICES=0,1,2,3 python train_vis.py --dataset simple --model DFF --backbone resnet50 --batch-size 8 --lr 0.001 --epochs 200 --crop-size 960 --kernel-size 5 --edge-weight 0.4
CUDA_VISIBLE_DEVICES=0,1,2,3 python train.py --dataset complex --model CASENet --backbone resnet152 --batch-size 4 --lr 0.001 --epochs 100 --crop-size 1280 --kernel-size 9 --edge-weight 0.4
CUDA_VISIBLE_DEVICES=0,1,2,3 python train_vis.py --dataset complex --model CASENet --backbone resnet152 --batch-size 4 --lr 0.001 --epochs 100 --crop-size 1280 --kernel-size 9 --edge-weight 0.4
CUDA_VISIBLE_DEVICES=0 python val.py --dataset simple --model DFF --backbone resnet50
CUDA_VISIBLE_DEVICES=0 python val.py --dataset complex --model CASENet --backbone resnet152
python val_mor.py
CUDA_VISIBLE_DEVICES=0 python test.py --dataset simple --model DFF --backbone resnet50
CUDA_VISIBLE_DEVICES=0 python test.py --dataset complex --model CASENet --backbone resnet152