Curved scene text detection via transverse and longitudinal sequence connection

• Published in: Pattern recognition Vol. 90; pp. 337 - 345
• Main Authors: Liu, Yuliang, Jin, Lianwen, Zhang, Shuaitao, Luo, Canjie, Zhang, Sheng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.06.2019
• Subjects: CNN; Curved dataset; Deep learning; LSTM; Scene text; Deep learning; Curved dataset; CNN; LSTM; Scene text
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2019.02.002

•We proposed a new curved dataset to facilitate curved detection method, whose annotation is based on relative objective method, which is very accurate.•We proposed a novel CTD method that can effectively detect both curved and non-curved text.•Seamless integration of a RNN method (TLOC) to significantly improve detection performance.•Implementation of polygonal post-processing methods (NPS and PNMS) to further improve results.•Our method achieves state-of-the-art performance on curved and non-curved datasets. Curved text detection is a difficult problem that has not been addressed sufficiently. To highlight the difficulties in reading curved text in a real environment, we constructed a curved text dataset called CTW1500, which includes over 10,000 text annotations in 1500 images, and used it to formulate a polygon-based curved text detector that can detect curved text without using an empirical combination. With the seamless integration of recurrent transverse and longitudinal offset connection, our method explores context information instead of predicting points independently, resulting in smoother and more accurate detection. Our approach is designed as a universal method, meaning it can be trained using rectangular or quadrilateral bounding boxes, requiring no extra effort. Experimental results on the CTW1500 dataset and Total-text demonstrated that our method with only a light backbone can outperform state-of-the-art methods by a large margin. Our method also achieved state-of-the-art performance on the MSRA-TD500 dataset, demonstrating its promising generalization ability. Code, datasets, and label-tool are available at https://github.com/Yuliang-Liu/Curve-Text-Detector.