for me the value is in improving on the algorithms for tasks like denoising, height map/normal map generation, light removal, edgedetection, tiling etc. These are things that can't be done perfectly with generalised standard algorithms as they require intuition about the content of the image. ML models can provide that…