Pluralistic-Inpainting

Why we included this project

Given one masked image, this CVPR 2019 implementation produces several distinct plausible fillings rather than a single guess, so you can sample different structures, colors, and textures for the missing region. That makes it a good fit for photo restoration, object removal, or creative editing where you want options instead of one answer. The repo bundles training, testing, and editing code, pretrained models for faces, buildings, and natural scenes, and a PyQt5 GUI for free-form masking. It's a research codebase from 2019, so expect older PyTorch dependencies and some manual setup before it runs.