search

taki0112 / UGATIT

Official Tensorflow implementation of U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation (ICLR 2020)
MIT License
6.17k stars 1.04k forks source link

Why trainA and trainB only have 3400 pictures? #65

Closed Z863058 closed 4 years ago

Z863058 commented 5 years ago

UGATIT is great! But I have some questions. 1.Why do you only use 3400 pictures in trainA and trainB? 2.if I use more pictures to train, for example 30000 trainA pictures and 30000 trainB pictures, need I change the iteration and epoch? 3.More trainA pictures is better, or more trainB pictures is better, or more trainA and trainB pictures is better? You are the best! Thank you!

Z863058 commented 5 years ago

And why don't you set iteration 3400 and epoch 300？

taki0112 commented 4 years ago
  1. I wanted to let you know that our model works well even though it's a small number
  2. I don't know. I think epoch 100 is enough.
  3. will be better