Reproduce “PATMAT: Person Aware Tuning of Mask-Aware Transformer for Face Inpainting” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Saman Motamed, Jianjin Xu, Chen Henry Wu, Fernando De la Torre
PATMAT adapts a pretrained face-inpainting model to preserve identity when filling masks. It tunes MAT using about 40 reference photos, fixing style anchors and manually grouping references by accessories or skin tone. In tests, PATMAT-C reached an ArcFace identity-similarity score of 0.66 versus 0.55 for MyStyle. It ranked third on FID, and results are limited to celebrity data and reference coverage.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.