SPTS v2: Single-Point Scene Text Spotting

• DOI: 10.48550/arxiv.2301.01635
• Publication Date: 2023-01-01
• Authors: Yuliang Liu; Jiaxin Zhang; Dezhi Peng; Mingxin Huang; Xinyu Wang; Jingqun Tang; Can Huang; Dahua Lin; Chunhua Shen; Xiang Bai; Lianwen Jin
• Keywords: Spotting; Representation; Sequence (biology)

End-to-end scene text spotting has made significant progress due to its intrinsic synergy between detection and recognition. Previous methods commonly regard manual annotations such as horizontal rectangles, rotated quadrangles, polygons a prerequisite, which are much more expensive than using single-point. Our new framework, SPTS v2, allows us train high-performing text-spotting models single-point annotation. v2 reserves the advantage of auto-regressive Transformer with an Instance Assignment Decoder (IAD) through sequentially predicting center points all instances inside same sequence, while Parallel Recognition (PRD) for recognition in parallel, significantly reduces requirement length sequence. These two decoders share parameters interactively connected simple but effective information transmission process pass gradient information. Comprehensive experiments on various existing benchmark datasets demonstrate can outperform previous state-of-the-art spotters fewer achieving 19$\times$ faster inference speed. Within context our suggest potential preference representation when compared other representations. Such attempt provides opportunity applications beyond realms paradigms. Code is available at: https://github.com/Yuliang-Liu/SPTSv2.