search

xh-liu / CC-FPSE

Code for NeurIPS 2019 paper "Learning to Predict Layout-to-image Conditional Convolutions for Semantic Image Synthesis"
128 stars 15 forks source link

How important is the perceptual loss? #8

Closed bearcatt closed 3 years ago

bearcatt commented 4 years ago

Hi @xh-liu, thanks for the awesome work. I found that the perceptual loss cannot decrease during training, which has also been found in original GauGAN. I am wondering whether this loss is useful or not? Have you tried not using the perceptual loss for your method?

Ha0Tang commented 4 years ago

@bearcatt Do you know how to train on 8 GPUs?

xh-liu commented 3 years ago

@bearcatt Yes, it is normal that the perceptual loss does not decrease during training. But it is important for training. Without the perceptual loss the model will have difficulty in convergence.