I think vector magic is the current state of the art: https://vectormagic.com/?=20 https://vectormagic.com/?=20 No one seems to have tried to leverage deep lea

I think vector magic is the current state of the art: https://vectormagic.com/?=20 https://vectormagic.com/?=20 No one seems to have tried to leverage deep learning yet; either because they haven't thought of doing so, or it just wouldn't be worthwhile. image to SVG's are an inherently deterministic task, with not much room for the noisy error of most deep learning models like stable diffusion and such. I think algorithmic approaches are just better in this case - deep learning isn't always superior.