I tried by myself but have some issues. Could you convert it to coreML and add to the gallery? https://github.com/CXH-Research/DocShadow-SD7K