UNETRSal: Saliency Prediction with Hybrid Transformer-Based Architecture
Résumé
Saliency prediction plays a critical role in understanding visual attention as it is a cornerstone for both natural scene understanding and automated document analysis. In this work, we propose the UN-ETRSal model for saliency prediction. Based on UNETR transformerbased model, we introduce a new decoder to increase efficiency on 2D images. Comprehensive evaluations on benchmark datasets, such as SAL-ICON and CAT2000, demonstrate that UNETRSal achieves state-ofthe-art performance across multiple saliency metrics, surpassing both conventional CNN-based and transformer-based methods. These results not only underscore the strengths of hybrid transformer architectures in modeling visual attention but also highlight the potential impact on advancing document representation modeling and layout analysis.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |