Feature Enhancement Network: A Refined Scene Text Detector

In this paper, we propose a refined scene text detector with a textitnovel Feature Enhancement Network (FEN) for Region Proposal and Text Detection Refinement. Retrospectively, both region proposal with textitonly 3× 3 sliding-window feature and text detection refinement with textitsingle scale high level feature are insufficient, especially for smaller scene text. Therefore, we design a new FEN network with textittask-specific, textitlow and textithigh level semantic features fusion to improve the performance of text detection. Besides, since textitunitary position-sensitive RoI pooling in general object detection is unreasonable for variable text regions, an textitadaptively weighted position-sensitive RoI pooling layer is devised for further enhancing the detecting accuracy. To tackle the textitsample-imbalance problem during the refinement stage, we also propose an effective textitpositives mining strategy for efficiently training our network. Experiments on ICDAR 2011 and 2013 robust text detection benchmarks demonstrate that our method can achieve state-of-the-art results, outperforming all reported methods in terms of F-measure.