peteryuX/esrgan-tf2

Issues

• Transfer learning

  #20 opened 3 years ago by eshwarkamal
  0

• Loss is always nan

  #4 opened 4 years ago by taoyu17
  1

• NOT an issue: It works amazingly well. However, use Spectral Normalization will converge VERY fast and nice !

  #15 opened 4 years ago by thusinh1969
  1

• Quality

  #16 opened 3 years ago by SolicTous
  0

• output SR images saved in comparison mode.

  #14 opened 4 years ago by santosh-shriyan
  0

• Speed

  #13 opened 4 years ago by SolicTous
  0

• loss_maybe_wrong

  #12 opened 4 years ago by weizhiliang0520
  0

• Use the LR_test image to test

  #11 opened 4 years ago by weizhiliang0520
  0

• How long did you training to get the pre_train model ckpt-500?

  #9 opened 4 years ago by weizhiliang0520
  0

• image with lower psnr , appearing visually good

  #2 opened 4 years ago by nanmehta
  7

• Esrgan_training D_loss=0

  #10 opened 4 years ago by weizhiliang0520
  0

• psnr traning_processing was stoped 'by itself' when step to 200000

  #8 opened 4 years ago by weizhiliang0520
  0

• building tfreacords file:

  #7 opened 4 years ago by weizhiliang0520
  1

• update the gradient do not need freeze one of network?

  #5 opened 4 years ago by cl886699
  2

• Why the training effect of using your own data set is very poor

  #1 opened 5 years ago by Haveoneriver
  4