Reproduce “TIRAD: Text-guided Image Residual Learning for Generalist Anomaly Detection” 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

Seongsu Lee, Seunghyun Oh, Seungjae Moon, Youngmin Ro

TIRAD uses text descriptions to filter visual differences in anomaly detection. It compares query patches with normal reference patches, subtracts a normal-text anchor from matched text representations, and uses the semantic residual to gate visual residuals. In four-shot transfer from MVTec AD to BraTS, it reports 83.80% image-level AUROC and 54.32% pixel-level F1, on one medical benchmark.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.