{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/detecting-curve-text-in-the-wild-new-dataset","title":"Detecting Curve Text in the Wild: New Dataset and New Solution","arxiv_id":"1712.02170","date":"2017-12-06","proceeding":null,"authors":["Liu Yuliang","Jin Lianwen","Zhang Shuaitao","Zhang Sheng"],"abstract":"Scene text detection has been made great progress in recent years. The\ndetection manners are evolving from axis-aligned rectangle to rotated rectangle\nand further to quadrangle. However, current datasets contain very little curve\ntext, which can be widely observed in scene images such as signboard, product\nname and so on. To raise the concerns of reading curve text in the wild, in\nthis paper, we construct a curve text dataset named CTW1500, which includes\nover 10k text annotations in 1,500 images (1000 for training and 500 for\ntesting). Based on this dataset, we pioneering propose a polygon based curve\ntext detector (CTD) which can directly detect curve text without empirical\ncombination. Moreover, by seamlessly integrating the recurrent transverse and\nlongitudinal offset connection (TLOC), the proposed method can be end-to-end\ntrainable to learn the inherent connection among the position offsets. This\nallows the CTD to explore context information instead of predicting points\nindependently, resulting in more smooth and accurate detection. We also propose\ntwo simple but effective post-processing methods named non-polygon suppress\n(NPS) and polygonal non-maximum suppression (PNMS) to further improve the\ndetection accuracy. Furthermore, the proposed approach in this paper is\ndesigned in an universal manner, which can also be trained with rectangular or\nquadrilateral bounding boxes without extra efforts. Experimental results on\nCTW-1500 demonstrate our method with only a light backbone can outperform\nstate-of-the-art methods with a large margin. By evaluating only in the curve\nor non-curve subset, the CTD + TLOC can still achieve the best results. Code is\navailable at https://github.com/Yuliang-Liu/Curve-Text-Detector.","url_abs":"http://arxiv.org/abs/1712.02170v1","url_pdf":"http://arxiv.org/pdf/1712.02170v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"detecting-curve-text-in-the-wild-new-dataset","repo_url":"https://github.com/Yuliang-Liu/Curve-Text-Detector","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"curved-text-detection","task_name":"Curved Text Detection"},{"task_slug":"scene-text-detection","task_name":"Scene Text Detection"},{"task_slug":"text-detection","task_name":"Text Detection"}],"methods":[],"datasets_introduced":[{"slug":"scut-ctw1500","name":"SCUT-CTW1500","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/curved-text-detection-on-scut-ctw1500","task":"Curved Text Detection","dataset":"SCUT-CTW1500","model":"CTD+TLOC","rank_in_archive_order":3,"of":5,"metrics":{"F-Measure":"73.4%"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1712.02170","atlas_url":"https://app.syntology.ai/?focus=1712.02170","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}