How did you train model specifically? And when fine-tuning on both UNet and VAE, did you also just use the single frame_50.png image to finetune?

johannakarras / DreamPose

Official implementation of "DreamPose: Fashion Image-to-Video Synthesis via Stable Diffusion"

MIT License

962 stars 73 forks source link

How did you train model specifically? And when fine-tuning on both UNet and VAE, did you also just use the single frame_50.png image to finetune? #32

Open zhangtao22 opened 1 year ago

zhangtao22 commented 1 year ago

I downloaded the base model and strictly followed the steps in your "Finetune on Sample" guide, but the resulting images are whitening and flickering.

johannakarras commented 1 year ago

I have recently tested and re-uploaded necessary finetuning scripts. The result is in media/demo.gif. Please let me know if these updates do not solve your issue!

zhangtao22 commented 1 year ago

@johannakarras The issue of whitening is solved.Nice!But the issue with wired movements still persists.