search

layumi / Seg-Uncertainty

IJCAI2020 & IJCV2021 :city_sunrise: Unsupervised Scene Adaptation with Memory Regularization in vivo
https://arxiv.org/abs/1912.11164
MIT License
387 stars 51 forks source link

Discriminator' Problem #10

Open gong-lei opened 4 years ago

gong-lei commented 4 years ago

Thank you for sharing this wonderful code first! And I have a small question in discriminator.I find the adversarial loss of AdaptSegNet is very unstable because of the global alignment in the segmentation output.I add non-local attention in the discriminator,but the performance drops dramatically.Then your discriminator,you say 'we follow the PatchGAN and deploy the multi-scale discrimimator model'.So for what consideration you utilize this strategy,and do you do an experiment to see how much performance improvement does this approach bring?

layumi commented 4 years ago

Hi @gong-lei Thanks for your attention. I do not compare the MS-Discriminator with Single-scale Discriminator.

The main reason is the experience in my previous work of CVPR19 (https://github.com/layumi/DGNet). Multi-scale discriminator works well in image generation, so I continue to use it.

Besides, the non-local operation is hard to tune. I also have tried it but the performance is not good.