ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks

1 Sep 2018Xintao WangKe YuShixiang WuJinjin GuYihao LiuChao DongChen Change LoyYu QiaoXiaoou Tang

The Super-Resolution Generative Adversarial Network (SRGAN) is a seminal work that is capable of generating realistic textures during single image super-resolution. However, the hallucinated details are often accompanied with unpleasant artifacts... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Image Super-Resolution BSD100 - 4x upscaling SRGAN + Residual-in-Residual Dense Block PSNR 27.85 # 4
SSIM 0.7455 # 7
Image Super-Resolution Manga109 - 4x upscaling bicubic PSNR 24.89 # 17
SSIM 0.7866 # 16
Image Super-Resolution Manga109 - 4x upscaling SRGAN + Residual-in-Residual Dense Block PSNR 31.66 # 4
SSIM 0.9196 # 5
Image Super-Resolution PIRM-test ESRGAN NIQE 2.55 # 2
Image Super-Resolution Set14 - 4x upscaling SRGAN + Residual-in-Residual Dense Block PSNR 28.99 # 5
SSIM 0.7917 # 7
Image Super-Resolution Set5 - 4x upscaling SRGAN + Residual-in-Residual Dense Block PSNR 32.73 # 3
SSIM 0.9011 # 7
Image Super-Resolution Urban100 - 4x upscaling SRGAN + Residual-in-Residual Dense Block PSNR 27.03 # 7
SSIM 0.8153 # 3
Image Super-Resolution Urban100 - 4x upscaling bicubic PSNR 23.14 # 33
SSIM 0.6577 # 29

Results from Other Papers


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK SOURCE PAPER COMPARE
Image Super-Resolution FFHQ 1024 x 1024 - 4x upscaling ESRGAN FID 72.73 # 8
MS-SSIM 0.782 # 8
PSNR 19.84 # 8
SSIM 0.353 # 8
Image Super-Resolution FFHQ 256 x 256 - 4x upscaling ESRGAN FID 166.36 # 8
MS-SSIM 0.747 # 8
PSNR 15.43 # 8
SSIM 0.267 # 8