Pixelwise loss in AAE - Githubissues

eriklindernoren / PyTorch-GAN

PyTorch implementations of Generative Adversarial Networks.

MIT License

16.22k stars 4.05k forks source link

Pixelwise loss in AAE #117

Open bhaswara opened 4 years ago

bhaswara commented 4 years ago

@eriklindernoren Is there any reasons why L1 loss is used here instead of using MSE loss for the pixelwise? And why it has to be multiplied with 0.999?

russab0 commented 3 years ago

The same question about Context Encoder

LucaLaFisca commented 1 year ago

I have the exact same question. Can you provide any answer?