ShichenLiu / SoftRas

Project page of paper "Soft Rasterizer: A Differentiable Renderer for Image-based 3D Reasoning"
MIT License
1.24k stars 156 forks source link

How can I add the texture of the 3D model in the code #87

Open M202071895 opened 3 years ago

M202071895 commented 3 years ago

The rendering process in the current code does not include textures, and the reconstruction basically only considers contours and shapes. How can I add the texture of the model, and whether the reconstruction effect will be better after adding it.

The dataset you are using is basically all black, which can only reflect the outline.

ShichenLiu commented 3 years ago

You can implement the texture reconstruction module following the network architecture shown in our paper (appendix). There are two ways of learning with textures:

(1): learn texture map (RGB): this won't affect the geometry. In fact, we fix the geometry and learn the texture on top of it in our implementation. (2): learn the "shading-like" color images: if the dataset contains shading images and the light source is known, the differentiable renderer can be seen as a shape-from-shading algorithm, thus improves the reconstruction. Our shading-aware model is also provided in the pretrained models (second bullet point after "We also provide our trained model here:")

hui1123 commented 2 years ago

Have you solved it? How to add texture to the generated mesh? @M202071895