Fitting Neural Textures and PBR Material Maps with ES (No Backprop)

A blog post from richg42 discusses how to use the ES (Efficiently Sampled) technique to fit neural textures and PBR (Physically-Based Rendering) material maps without the need for backpropagation. The technique uses a forward pass to compute the gradient of the loss function, allowing for more efficient training of neural networks. This approach is particularly useful for tasks such as material mapping and texture synthesis. The post includes code examples in PyTorch and TensorFlow to demonstrate the technique.