pretrained_models

Place pretrained models here.

Pretrained models

1. SRResNet_bicx4_in3nf64nb16.pth: the well-trained SRResNet model in PSNR orientation.
2. SRGAN.pth: the pretrained model SRGAN implemented by BasicSR.
3. ESRGAN_SuperSR.pth: the pretrained model ESRGAN_SuperSR.

Three pretrained Ranker models :

1. Ranker_NIQE.pth: the well-trained Ranker with NIQE metric.
2. Ranker_Ma.pth: the well-trained Ranker with Ma metric.
3. Ranker_PI.pth: the well-trained Ranker with PI metric.

RankSRGAN models

1. RankSRGAN_NIQE.pth: the RankSRGAN in NIQE orientation.
2. RankSRGAN_Ma.pth: the RankSRGAN in Ma orientation.
3. RankSRGAN_PI.pth: the RankSRGAN in PI orientation.

Note that the pretrained models are trained under the MATLAB bicubic kernel. If the downsampled kernel is different from that, the results may have artifacts.