TextFormer: A Query-based End-to-End Text Spotter with Mixed Supervision

TL;DR

TextFormer introduces a Transformer-based, query-driven end-to-end framework for text detection and recognition that leverages mixed supervision and a novel global aggregation module, achieving superior performance on multilingual benchmarks.

Contribution

The paper presents a novel query-based Transformer architecture with an Adaptive Global Aggregation module and mixed supervision, enhancing end-to-end text spotting performance.

Findings

Outperforms state-of-the-art on TDA-ReCTS with 13.2% improvement in 1-NED

Effectively integrates detection and recognition with shared features

Utilizes mixed supervision to improve detection and recognition accuracy

Abstract

End-to-end text spotting is a vital computer vision task that aims to integrate scene text detection and recognition into a unified framework. Typical methods heavily rely on Region-of-Interest (RoI) operations to extract local features and complex post-processing steps to produce final predictions. To address these limitations, we propose TextFormer, a query-based end-to-end text spotter with Transformer architecture. Specifically, using query embedding per text instance, TextFormer builds upon an image encoder and a text decoder to learn a joint semantic understanding for multi-task modeling. It allows for mutual training and optimization of classification, segmentation, and recognition branches, resulting in deeper feature sharing without sacrificing flexibility or simplicity. Additionally, we design an Adaptive Global aGgregation (AGG) module to transfer global features into sequential features for reading arbitrarily-shaped texts, which overcomes the sub-optimization problem of RoI operations. Furthermore, potential corpus information is utilized from weak annotations to full labels through mixed supervision, further improving text detection and end-to-end text spotting results. Extensive experiments on various bilingual (i.e., English and Chinese) benchmarks demonstrate the superiority of our method. Especially on TDA-ReCTS dataset, TextFormer surpasses the state-of-the-art method in terms of 1-NED by 13.2%.