Put the pth files in the folder "model".
- Test
python test.py --model=usbnet --cs_ratio=25
The results will be generated in the folder "./results/usbnet/{dataset}/{cs_ratio}/", where results.csv will save the results in the format "{Image},{PSNR},{SSIM},{Time}".
- Train
- Multi-GPU
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 train.py --model=usbnet --data_path="" --eval_data_path="" --cs_ratio=10 --blr=1e-4 --min_lr=1e-6 --epochs=400 --batch_size=16 --warmup_epochs=10 --input_size=96
- Single GPU
python train.py --model=usbnet --data_path="" --eval_data_path="" --cs_ratio=10 --blr=1e-4 --min_lr=1e-6 --epochs=400 --batch_size=16 --warmup_epochs=10 --input_size=96
Put the pth files in the folder "model-Cartesian" or "model-Radial".
- Test
python test.py --model=usbnet --cs_ratio=5 --input_size=256 --mask_type=Radial
The results will be generated in the folder "./results/usbnet/{dataset}/{cs_ratio}/", where results.csv will save the results in the format "{Image},{PSNR},{SSIM},{Time}".
- Train
- Multi-GPU
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 train.py --model=usbnet --data_path="" --eval_data_path="" --cs_ratio=5 --blr=5e-5 --min_lr=1e-6 --epochs=100 --batch_size=1 --warmup_epochs=10 --input_size=256 --mask_type=Radial
- Single GPU
python train.py --model=usbnet --data_path="" --eval_data_path="" --cs_ratio=5 --blr=5e-5 --min_lr=1e-6 --epochs=100 --batch_size=1 --warmup_epochs=10 --input_size=256 --mask_type=Radial
Put the pth files in the folder "Sim".
python test.py --data_root="path of data"
Put the pth files in the folder "Real".
python test_real.py --data_path="path of data" --mask_path="path of mask"
-
Model for ICS: [BaiduYun] or [GoogleDrive]
-
Model for CS-MRI: [BaiduYun] or [GoogleDrive]/[GoogleDrive]
-
Model for SCI: [BaiduYun]