Giters

CVMI-Lab / Point-UV-Diffusion

(ICCV2023) This is the official PyTorch implementation of ICCV2023 paper: Texture Generation on 3D Meshes with Point-UV Diffusion

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Code for conditional texture generation

yuyingyeh opened this issue · comments

commented

Hi,

Thanks for your work! I found that you also demonstrate conditional texture generation in your paper. I'm interested in trying your method of texture generation for meshes conditioned on image using user-provided data.

For the current release, it seems that there is no instruction on how to evaluate on user-provided (image, mesh) pair. Would you also release your code for this part? Thank you very much!

commented

Hi,

Thanks for your work! I found that you also demonstrate conditional texture generation in your paper. I'm interested in trying your method of texture generation for meshes conditioned on image using user-provided data.

For the current release, it seems that there is no instruction on how to evaluate on user-provided (image, mesh) pair. Would you also release your code for this part? Thank you very much!

The code we've released currently handles images from the processed ShapeNet dataset and uses pre-extracted CLIP features as input. To adapt this for custom images and meshes for texture generation, the following steps are needed:

  1. Load the image and extract the CLIP feature.
  2. Load the mesh and use xatlas for UV-mapping, also calculating the corresponding XYZ, normal map.

I might think about creating a script for this in the future, but I don't have such plans right now, and :) I can't make any promises. In the meantime, you might want to try preparing your data following these steps on your own.

Feel free to reach out if you have more questions!

commented

Hi, please refer to this issue

commented

Hi,

Thanks for your work! I found that you also demonstrate conditional texture generation in your paper. I'm interested in trying your method of texture generation for meshes conditioned on image using user-provided data.

For the current release, it seems that there is no instruction on how to evaluate on user-provided (image, mesh) pair. Would you also release your code for this part? Thank you very much!

Hello, have you solved this problem? Can you share the code for this content?