hello451 comments

Repositories
Issues
Comments

Results 3 comments of


                                            hello451

How can I generate images from my own captions?

Same questions..

Without feature distillation, it can get even better results?

I have got similar situation... So maybe the most important thing is the classifiers from shallow to deep

I trained on UBC Fashion Dataset，but the outcome is pretty bad.

Just find a good [repo](https://github.com/Wangt-CN/DisCo) for similar task. They got much better results compared to DreamPose and quantitavely compare with DreamPose in their paper.