Text-DIAE: A Self-Supervised Degradation Invariant Autoencoder for Text Recognition and Document Enhancement

• DOI: 10.1609/aaai.v37i2.25328
• Publication Date: 2023-06-27T16:13:03Z
• Authors: Mohamed Ali Souibgui; Sanket Biswas; Andres Mafla; Ali Furkan Biten; Alicia Fornés; Yousri Kessentini; Josep Lladós; Lluis Gomez; Dimosthenis Karatzas
• Keywords: Pretext; Autoencoder; Text recognition

In this paper, we propose a Text-Degradation Invariant Auto Encoder (Text-DIAE), self-supervised model designed to tackle two tasks, text recognition (handwritten or scene-text) and document image enhancement. We start by employing transformer-based architecture that incorporates three pretext tasks as learning objectives be optimized during pre-training without the usage of labelled data. Each is specifically tailored for final downstream tasks. conduct several ablation experiments confirm design choice selected Importantly, proposed does not exhibit limitations previous state-of-the-art methods based on contrastive losses, while at same time requiring substantially fewer data samples converge. Finally, demonstrate our method surpasses in existing supervised settings handwritten scene Our code trained models will made publicly available https://github.com/dali92002/SSL-OCR