search

xinntao / ESRGAN

ECCV18 Workshops - Enhanced SRGAN. Champion PIRM Challenge on Perceptual Super-Resolution. The training codes are in BasicSR.
https://github.com/xinntao/BasicSR
Apache License 2.0
5.91k stars 1.05k forks source link

what is difference between RRDB_ESRGAN_x4 and RRDB_PSNR_x4 #85

Open suke27 opened 4 years ago

kgorins commented 4 years ago

As far as I understood, RRDB_ESRGAN_x4 is the current ESRGAN model, and RRDB_PSNR_x4 is a standart SRGAN one that gives smoother results (can be found here https://github.com/xinntao/BasicSR/blob/master/metrics/calculate_PSNR_SSIM.py).

victorca25 commented 4 years ago

No, PSNR is the model trained only using content loss to initialize the filters. You use the PSNR model as a pre-trained model to train the ESRGAN model with feature loss and adversarial loss, so it converges faster to the global minima. Read the paper, it's explained there.