search

SiTH-Diffusion / SiTH

[CVPR 2024] SiTH: Single-view Textured Human Reconstruction with Image-Conditioned Diffusion
https://ait.ethz.ch/sith
MIT License
120 stars 10 forks source link

About the diffusion model #4

Closed laugher07 closed 4 months ago

laugher07 commented 4 months ago

Amazing work! I will appreciate it if you can tell me how to train the diffusion model (hallucinate) or give me some advice on how to learn and start the work(about diffusion model) like this?

azuxmioy commented 4 months ago

Hi,

Sorry for the delay. We are still refactoring the training code to fit the new dataset format. We plan to release them near the main conference, however our recent schedule is a bit tight.

We can send you the old training scripts (not cleaned and tested yet), but you will need to adapt the data loader. Please send an email to custom.humans@gmail.com

Thanks