Irregular text block recognition via decoupling visual, linguistic, and positional information

Ziyan Li, Lianwen Jin, Chengquan Zhang, Jiaxin Zhang, Zecheng Xie, Pengyuan Lyu, Kun Yao

Abstract: Highlights•Proposing a new challenging task called irregular text block recognition (ITBR).•Designing a progressive layout reasoning network (PLRN) to address ITBR problem.•Two key modules for information decoupling: character spotting and linkage reasoning.•Constructing a new TB600 dataset with 600 text block images for benchmarking.•PLRN achieves superior performance over previous state-of-the-art methods on TB600.