search

microsoft / StyleSwin

[CVPR 2022] StyleSwin: Transformer-based GAN for High-resolution Image Generation
https://arxiv.org/abs/2112.10762
MIT License
503 stars 49 forks source link

pre_training and fine-tune #13

Closed Joker-ZXR closed 2 years ago

Joker-ZXR commented 2 years ago

Hi, thank you so much for sharing your code！ I have a question. I'd like to transfer the model to medical images. But I don't I retrain directly or I need load the pre-training model and fine-tune it? What do you think about it? Looking forward to your reply!

ForeverFancy commented 2 years ago

Thanks for your interest of our work. In my opinion, there might be a large gap between the distribution of medical images and our pretrained datasets. As a result, I think it might be more appropriate to retrain our model using your dataset. However, you could also try to fine-tune the model to see if it is feasible.

Joker-ZXR commented 2 years ago

Ok, thank you for your answer. I'll try it.