svg to mask?

[GDbbq]

Hello, I would like to ask how to generate a semantic segmentation mask through SVG? Can I provide the docker environment, I used the environment I built to train and reported errors, and I couldn't successfully build the VIT model, which may be a problem with the timm version。