CLEAR: Cross-Transformers With Pre-Trained Language Model for Person Attribute Recognition and Retrieval

Doanh C. Bui, Thinh V. Le, Ba Hung Ngo, Tae Jong Choi

Abstract: Highlights•We ran PAR tests with robust C2<math><msup is="true"><mrow is="true"></mrow><mrow is="true"><mn is="true">2</mn></mrow></msup></math>T-Net using local & global dependencies.•Leverage GPT embeddings to create pseudo text for key attribute queries in AR tasks.•We propose a training strategy adapting C2<math><msup is="true"><mrow is="true"></mrow><mrow is="true"><mn is="true">2</mn></mrow></msup></math>T-Net from PAR to AR, CLEAR.•CLEAR achieves SOTA on PAR and AR across PA100K, PETA, RAPv2, Market-1501 and UPAR.