Sequential Visual and Semantic Consistency for Semi-supervised Text Recognition

arXiv:2402.15806v1 Announce Type: new
Abstract: Scene text recognition (STR) is a challenging task that requires large-scale annotated data for training. However, collecting and labeling real text images is expensive and time-consuming, which limits the availability of real data. Therefore, most existing STR methods resort to synthetic data, which may introduce domain discrepancy and degrade the performance of STR models. To alleviate this problem, recent semi-supervised STR methods exploit unlabeled real data by enforcing character-level consistency regularization between weakly and strongly …

abstract annotated data arxiv availability cs.cv data domain images labeling real data recognition scale semantic semi-supervised synthetic synthetic data text training type visual