About the loss

[SongYxing]

I want to know what l_d_fake and l_d_real mean, l_g_gan is the sum of discriminator and generator? And what the 'd' and 'g' in l_d_fake and l_g_pix_s stand for?

[IanYeung]

I want to know what l_d_fake and l_d_real mean, l_g_gan is the sum of discriminator and generator? And what the 'd' and 'g' in l_d_fake and l_g_pix_s stand for?

l_d_fake and l_d_real are losses for the discriminator. d means discriminator and g means generator.

[SongYxing]

real is GT , fake is Y_predict?

[IanYeung]

real is GT , fake is Y_predict?

Yes.

[SongYxing]

I used your dataset to train on train_EDVR-GAN_woTSA_RealVSR_YCbCr_Split, and found that psnr_avg has been gradually decreasing, and finally dropped to 25.44. Shouldn't psnr be on the rise?

[IanYeung]

I used your dataset to train on train_EDVR-GAN_woTSA_RealVSR_YCbCr_Split, and found that psnr_avg has been gradually decreasing, and finally dropped to 25.44. Shouldn't psnr be on the rise?

If you use GAN in training, the psnr will decrease.

[SongYxing]

Could u tell me why？

[IanYeung]

Could u tell me why？

GAN training is usually for visual quality but not for objective measures like psnr and ssim.

[SongYxing]

if 'g' is generator,what is l_g_gan? Doesn't gan mean GAN?

[IanYeung]

if 'g' is generator,what is l_g_gan? Doesn't gan mean GAN?

Yes it is the GAN loss for generator.