search

zju3dv / RelightableAvatar

[CVPR 2024 (Highlight)] Relightable and Animatable Neural Avatar from Sparse-View Video
https://zju3dv.github.io/relightable_avatar/
95 stars 3 forks source link

Instructions on training #1

Open lizhe00 opened 5 months ago

lizhe00 commented 5 months ago

Hi, nice work! Could you please tell me when to provide instructions on training?

HelloRicky123 commented 4 months ago

Hi Sida, I have the same question. Do you have plan for the two TODO list recently? @pengsida

pengsida commented 4 months ago

Hi, do you need it urgently?

HelloRicky123 commented 4 months ago

Actually yes, a few days ago, I intended to incorporate this code into my experiment and submit it to Siggraph Asia. However, I encountered issues during the training process and when using the previous SyntheticHuman dataset to evaluate the lighting effects, resulting in a failed submission. Now, I aim to continue to enhance the experiment using this test data and evaluation criteria.

pengsida commented 4 months ago

Our code for the training stage has already been open-sourced, but we haven't written the documents yet. We will find time to write them in the next few weeks.

HelloRicky123 commented 4 months ago

Yes, I have tested the training code and found it to be beneficial. I am looking forward to receiving the document. Could you please share the SyntheticHuman++ dataset with me? Additionally, I would like the file for evaluating Albedo and Relighting in the paper.

Thanks a lot!

pengsida commented 4 months ago

Here are the Mobile-Stage dataset and SyntheticHuman++ dataset.

HelloRicky123 commented 4 months ago

Received. Thanks a lot again!

JiatengLiu commented 2 months ago

Hi, @pengsida @HelloRicky123 . Thanks for your share, but I can not unzip the SyntheticHuman++ dataset correctly, could you provide the instruction?