{"id":18322532,"url":"https://github.com/tencentarc/bts","last_synced_at":"2026-02-08T15:01:25.882Z","repository":{"id":109180073,"uuid":"467754442","full_name":"TencentARC/BTS","owner":"TencentARC","description":"BTS: A Bi-lingual Benchmark for Text Segmentation in the Wild","archived":false,"fork":false,"pushed_at":"2024-04-16T07:11:21.000Z","size":2708,"stargazers_count":31,"open_issues_count":5,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-07-21T06:49:43.502Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/TencentARC.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.md","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2022-03-09T02:49:32.000Z","updated_at":"2025-07-14T08:48:51.000Z","dependencies_parsed_at":"2024-11-05T18:44:39.566Z","dependency_job_id":null,"html_url":"https://github.com/TencentARC/BTS","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/TencentARC/BTS","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TencentARC%2FBTS","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TencentARC%2FBTS/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TencentARC%2FBTS/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TencentARC%2FBTS/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/TencentARC","download_url":"https://codeload.github.com/TencentARC/BTS/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/TencentARC%2FBTS/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29234154,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-08T14:18:14.570Z","status":"ssl_error","status_checked_at":"2026-02-08T14:18:14.071Z","response_time":57,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-11-05T18:25:02.826Z","updated_at":"2026-02-08T15:01:25.858Z","avatar_url":"https://github.com/TencentARC.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# BTS: A Bi-lingual Benchmark for Text Segmentation in the Wild\n\nThis is the repo to host the dataset BTS from the following paper:\n\n[Xixi Xu](), [Zhongang Qi](), [Jianqi Ma](https://github.com/mjq11302010044/), [Honglun Zhang](), [Ying Shan](), [Xiaohu Qie](), **BTS: A Bi-lingual Benchmark for Text Segmentation in the Wild**\n\nSummary of license permissions：\n\n**Our dataset is now fully released for academic use.** \nThe researcher shall use the BTS dataset only for non-commercial algorithm research and educational purposes. Except for the above purposes, the researcher may not use the BTS dataset for any other purposes, including but not limited to distribution, commercial use, advertising, etc.\n\nYou can download the dataset from the following link, only if you agree to the above permissions.\n\nhttps://drive.weixin.qq.com/s?k=AJEAIQdfAAofh5N4rQ\n\n\n\n## Selection of scenes. \nThe key motivation of the selection of scenes is to ensure the representation and generalization of the dataset. \n- First, we have images indoor and outdoor to balance the lighting conditions. \n- Second, the text line appearance variety is also an important factor to be considered, i.e., \ntext line in different orientation (vertical and horizontal text in couplets and textbooks) and \ncurve-shaped (some of the signboards). \n- The third factor lies in the font diversity, e.g., \nwe have text images in printed font in textbook and artistic font on the signboard. \n\nWe believe that varieties in these three perspectives can ensure the segmentation model \nto be well-trained with better generalization.\n\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"./figure/BTS_examples.png\" width=\"70%\"\u003e\n\u003c/p\u003e\n\n## Dataset annotation.\nBTS eliminate algorithms and out-of-the-box models for the labeling process to prevent some bad labeling cases. \nThe annotation workflow is as follows.\n\n- Images cleaning. Unqualified examples such as fuzzy images with unrecognizable characters and strokes will be filtered out.\n- Manual annotation. All the images in BTS are manually annotated by humans in three levels, \nincluding the pixel-level, the character-level, and the line-level annotations. \nPhotoShop is the main tool. The pencil tool in Photoshop is utilized to assist the annotators \nto label pixel-level mask annotations for texts.  \n- Two rounds of quality checks. \nDuring the labeling process, annotators will cross check the annotations from each other; \nafter the labeling process, several professional researchers will double check the annotations. \n\nThe designed workflow ensures all annotations to be made in relatively high quality and benchmark to be highly-reliable.\n\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"./figure/anno.bmp\" width=\"70%\"\u003e\n\u003c/p\u003e\n\n## Dataset statistics.\nBTS contains 14250 images. \n\n\u003cp align=\"center\"\u003e\n  \u003cimg src=\"./figure/dis_scene_pic.png\" width=\"70%\"\u003e\n\u003c/p\u003e\n\nThe distribution is nearly balanced,\nwhich is consistent with real-world distribution.\n\n## Download\n\nA full download should contain these files:\n\n* ```BTS_VAL.zip``` contains 10,188 images.\n* ```BTS_TRAIN.zip``` contains 2,696 images.\n* ```BTS_TEST.zip``` contains 1,366 images.\n\nIn each zip packages, there are three folds.\n* ```image``` contains original images.  ```[SceneID]_[SampleID].jpg``` \n\n* ```bpoly_label``` word-level and char-level labels corresponding to the images.   ```[SceneID]_[SampleID]_anno.json``` \n\n* ```semantic_label``` mask labels corresponding to the images.   ```[SceneID]_[SampleID]_maskfg.png``` \n\nIn this table, we compare BTS with a variety of representative datasets. \n\n\u003cfont size=\"11\" face=\"Courier New\"\u003e\n\u003ctable\u003e\n  \u003ctr\u003e\n    \u003ctd colspan=\"1\"\u003eDataset\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eText Type\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eImages\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eWords\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eChars\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eMasks\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eChar Classes\u003c/td\u003e\n    \u003ctd colspan=\"1\"\u003eLanguage\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eICDAR13 FST\u003c/td\u003e\n    \u003ctd\u003eScene\u003c/td\u003e\n    \u003ctd\u003e462\u003c/td\u003e\n    \u003ctd\u003e1944\u003c/td\u003e\n    \u003ctd\u003e6620\u003c/td\u003e\n    \u003ctd\u003eWord,Char\u003c/td\u003e\n    \u003ctd\u003e36\u003c/td\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eCOCO_TS\u003c/td\u003e\n    \u003ctd\u003eScene\u003c/td\u003e\n    \u003ctd\u003e14690\u003c/td\u003e\n    \u003ctd\u003e139034\u003c/td\u003e\n    \u003ctd\u003e-\u003c/td\u003e\n    \u003ctd\u003eWord\u003c/td\u003e\n    \u003ctd\u003e36\u003c/td\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eMLT_S\u003c/td\u003e\n    \u003ctd\u003eScene\u003c/td\u003e\n    \u003ctd\u003e6896\u003c/td\u003e\n    \u003ctd\u003e30691\u003c/td\u003e\n    \u003ctd\u003e-\u003c/td\u003e\n    \u003ctd\u003eWord\u003c/td\u003e\n    \u003ctd\u003e36\u003c/td\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eTotal-Text\u003c/td\u003e\n    \u003ctd\u003eScene\u003c/td\u003e\n    \u003ctd\u003e1555\u003c/td\u003e\n    \u003ctd\u003e9330\u003c/td\u003e\n    \u003ctd\u003e-\u003c/td\u003e\n    \u003ctd\u003eWord\u003c/td\u003e\n    \u003ctd\u003e36\u003c/td\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003eTextSeg\u003c/td\u003e\n    \u003ctd\u003eScene+Design\u003c/td\u003e\n    \u003ctd\u003e4024\u003c/td\u003e\n    \u003ctd\u003e15691\u003c/td\u003e\n    \u003ctd\u003e73790\u003c/td\u003e\n    \u003ctd\u003eWord,Word-Effect,Char\u003c/td\u003e\n    \u003ctd\u003e36\u003c/td\u003e\n    \u003ctd\u003eEnglish\u003c/td\u003e\n  \u003c/tr\u003e\n  \u003ctr\u003e\n    \u003ctd\u003e\u003cb\u003eBTS(Ours)\u003c/b\u003e\u003c/td\u003e\n    \u003ctd\u003eScene\u003c/td\u003e\n    \u003ctd\u003e14250\u003c/td\u003e\n    \u003ctd\u003e44280\u003c/td\u003e\n    \u003ctd\u003e209090\u003c/td\u003e\n    \u003ctd\u003eWord,Char\u003c/td\u003e\n    \u003ctd\u003e3985\u003c/td\u003e\n    \u003ctd\u003eBi-lingual\u003c/td\u003e\n  \u003c/tr\u003e\n\u003c/table\u003e\n\u003c/font\u003e\n\n\n## Acknowledgements\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftencentarc%2Fbts","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftencentarc%2Fbts","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftencentarc%2Fbts/lists"}