search

thuml / Transfer-Learning-Library

Transfer Learning Library for Domain Adaptation, Task Adaptation, and Domain Generalization
http://transfer.thuml.ai
MIT License
3.39k stars 553 forks source link

about the generated image problem #217

Closed KOOKOKOK closed 1 year ago

KOOKOKOK commented 1 year ago

Hi,when i run examples/domain_adaptation/semantic_segmentationcycada.py, 00448

threre is the phantom on the top of some genareted images.how can i eliminate them?

thucbx99 commented 1 year ago

This is because the generator of CycleGAN fails to capture the complex scene in the picture. We also find this issue when performing translation on other datasets. So the problem is basically, how to train the generator (with limited amount of data & relative small generator). Maybe tuning some of hyper-parameters can be better. But I think this is an open problem for GAN now.

KOOKOKOK commented 1 year ago

thanks for your answer,i will try to improving💕