All You Need Is Boundary: Toward Arbitrary-Shaped Text Spotting

About

Recently, end-to-end text spotting that aims to detect and recognize text from cluttered images simultaneously has received particularly growing interest in computer vision. Different from the existing approaches that formulate text detection as bounding box extraction or instance segmentation, we localize a set of points on the boundary of each text instance. With the representation of such boundary points, we establish a simple yet effective scheme for end-to-end text spotting, which can read the text of arbitrary shapes. Experiments on three challenging datasets, including ICDAR2015, TotalText and COCO-Text demonstrate that the proposed method consistently surpasses the state-of-the-art in both scene text detection and end-to-end text recognition tasks.

Hao Wang, Pu Lu, Hui Zhang, Mingkun Yang, Xiang Bai, Yongchao Xu, Mengchao He, Yongpan Wang, Wenyu Liu · 2019

Related benchmarks

Task	Dataset	Result
Text Detection	ICDAR 2015	Precision89.8	188
Text Detection	Total-Text	Precision88.9	160
Scene Text Detection	ICDAR 2015 (test)	F1 Score88.6	150
Text Detection	Total-Text (test)	F-Measure87	126
Text Detection	ICDAR 2015 (test)	F1 Score88.6	108
Scene Text Detection	TotalText (test)	Recall85	106
Scene Text Spotting	Total-Text (test)	F-measure (None)65	105
End-to-End Text Spotting	ICDAR 2015	Strong Score79.7	104
Scene Text Detection	Total-Text	Precision85.2	89
End-to-End Scene Text Spotting	Total-Text	Hmean (None)67.8	80