monyone / SISR

Single Image Super Resolution (SISR) implementation by PyTorch

Single Image Super Resolution (SISR)

Requirement

• python3
• pytorch
• torchvision

Models

• SRCNN (ECCV 2014) [Paper]
• FSRCNN (ECCV 2016) [Paper]
• ESPCN (CVPR 2016) [Paper]
• VDSR (CVPR 2016) [Paper]
• DRCN (CVPR Oral 2016) [Paper]
• RED-Net (CVPR 2016) [Paper]
• DRRN (CVPR 2017) [Paper]
• LapSRN (CVPR 2017) [Paper]
• SRResNet, SRGAN (CVPR Oral 2017) [Paper]
• EDSR (CVPR Workshop 2017) [paper]
• EnhanceNet (ICCV 2017) [Paper]
• SRDenseNet (ICCV 2017) [Paper]
• MSLapSRN (TPAMI 2018) [Paper]
• TSRN (ECCV 2018) [Paper]
• RDN (CVPR 2018) [Paper]
• RRDBNet, ESRGAN (ECCV 2018) [Paper]
• RCAN (ECCV 2018) [Paper]
• SAN (CVPR 2019) [Paper]
• Swift-SRGAN (2021) [Paper]
• Real-ESRNet, Real-ESRGAN (ICCV Workshop 2021) [Paper]
• HPUN (AAAI 2023) [Paper]

Usage

Train

Geneator

./train.py --generator [ONE_OF_ABOVE_MODELS] --epoch ${epochs} --scale ${upscale_factor}

GAN

./train.py --generator [ONE_OF_ABOVE_MODELS] --discriminator [ONE_OF_ABOVE_DISCRIMINATIOR] --epoch ${epochs} --scale ${upscale_factor} --distort [Real-ESRGAN|BSRGAN|JPEG]

Testing

./main.py --model [ONE_OF_ABOVE_MODELS] --image ${image_path} --state ${state_path} --scale ${upscale_factor} --test --distort [Real-ESRGAN|BSRGAN|JPEG]

Super Resolution

./main.py --model [ONE_OF_ABOVE_MODELS] --image ${image_path} --state ${state_path} --scale ${upscale_factor}