{"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/textcaps-handwritten-character-recognition","title":"TextCaps : Handwritten Character Recognition with Very Small Datasets","arxiv_id":"1904.08095","date":"2019-04-17","proceeding":null,"authors":["Vinoj Jayasundara","Sandaru Jayasekara","Hirunima Jayasekara","Jathushan Rajasegaran","Suranga Seneviratne","Ranga Rodrigo"],"abstract":"Many localized languages struggle to reap the benefits of recent advancements\nin character recognition systems due to the lack of substantial amount of\nlabeled training data. This is due to the difficulty in generating large\namounts of labeled data for such languages and inability of deep learning\ntechniques to properly learn from small number of training samples. We solve\nthis problem by introducing a technique of generating new training samples from\nthe existing samples, with realistic augmentations which reflect actual\nvariations that are present in human hand writing, by adding random controlled\nnoise to their corresponding instantiation parameters. Our results with a mere\n200 training samples per class surpass existing character recognition results\nin the EMNIST-letter dataset while achieving the existing results in the three\ndatasets: EMNIST-balanced, EMNIST-digits, and MNIST. We also develop a strategy\nto effectively use a combination of loss functions to improve reconstructions.\nOur system is useful in character recognition for localized languages that lack\nmuch labeled training data and even in other related more general contexts such\nas object recognition.","url_abs":"http://arxiv.org/abs/1904.08095v1","url_pdf":"http://arxiv.org/pdf/1904.08095v1.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":"textcaps-handwritten-character-recognition","repo_url":"https://github.com/vinojjayasundara/textcaps","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"textcaps-handwritten-character-recognition","repo_url":"https://github.com/kubantjan/fast-form","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"textcaps-handwritten-character-recognition","repo_url":"https://github.com/milanzongor/unihack_2020","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"few-shot-image-classification","task_name":"Few-Shot Image Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-generation","task_name":"Image Generation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-classification-on-emnist-letters","task":"Image Classification","dataset":"EMNIST-Letters","model":"TextCaps","rank_in_archive_order":6,"of":11,"metrics":{"Accuracy":"95.39"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-fashion-mnist","task":"Image Classification","dataset":"Fashion-MNIST","model":"TextCaps","rank_in_archive_order":8,"of":34,"metrics":{"Percentage error":"6.29"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-mnist","task":"Image Classification","dataset":"MNIST","model":"TextCaps","rank_in_archive_order":15,"of":81,"metrics":{"Accuracy":"99.71","Percentage error":"0.29"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}