Fitting Neural Textures and PBR Material Maps with ES (No Backprop)
A new C++ experiment on GitHub demonstrates a method for fitting neural textures and PBR material maps using Evolution Strategies instead of traditional backpropagation. The approach uses a low-resolution latent texture and a small MLP decoder to achieve efficient image reconstruction.
Why it matters
This research offers an alternative to gradient-based training for neural rendering, potentially simplifying the computational requirements for specific graphics tasks.
The repo, released on September 4, 2026 is here:
https://github.com/richgel999/neural_texture_es2
Here is a mirror of the README.md file on GitHub, with the Prior Art disclosure.
A small, self-contained C++ experiment: an RGB image (or up to four same-size RGB textures of one material) is encoded as a shared low-resolution latent texture plus a tiny MLP decoder, and both are trained entirely with Evolution Strategies — no backprop, no autodiff, no training framework. (An optional late-training polish, --mlp-fd , switches the decoder to numerical finite differences; still no backprop.) Dependencies are stb_image , stb_image_write , and OpenMP.
Write-up: Fitting a neural texture decoder with ES
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in