{"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/dtrocr-decoder-only-transformer-for-optical","title":"DTrOCR: Decoder-only Transformer for Optical Character Recognition","arxiv_id":"2308.15996","date":"2023-08-30","proceeding":null,"authors":["Masato Fujitake"],"abstract":"Typical text recognition methods rely on an encoder-decoder structure, in which the encoder extracts features from an image, and the decoder produces recognized text from these features. In this study, we propose a simpler and more effective method for text recognition, known as the Decoder-only Transformer for Optical Character Recognition (DTrOCR). This method uses a decoder-only Transformer to take advantage of a generative language model that is pre-trained on a large corpus. We examined whether a generative language model that has been successful in natural language processing can also be effective for text recognition in computer vision. Our experiments demonstrated that DTrOCR outperforms current state-of-the-art methods by a large margin in the recognition of printed, handwritten, and scene text in both English and Chinese.","url_abs":"https://arxiv.org/abs/2308.15996v1","url_pdf":"https://arxiv.org/pdf/2308.15996v1.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":"dtrocr-decoder-only-transformer-for-optical","repo_url":"https://github.com/arvindrajan92/DTrOCR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"handwritten-text-recognition","task_name":"Handwritten Text Recognition"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"optical-character-recognition","task_name":"Optical Character Recognition"},{"task_slug":"optical-character-recognition","task_name":"Optical Character Recognition (OCR)"},{"task_slug":"scene-text-recognition","task_name":"Scene Text Recognition"},{"task_slug":"task-2","task_name":"Task 2"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/handwritten-text-recognition-on-iam","task":"Handwritten Text Recognition","dataset":"IAM","model":"DTrOCR 105M","rank_in_archive_order":1,"of":17,"metrics":{"CER":"2.38"},"uses_additional_data":false},{"leaderboard":"/sota/optical-character-recognition-on-benchmarking","task":"Optical Character Recognition (OCR)","dataset":"Benchmarking Chinese Text Recognition: Datasets, Baselines, and an Empirical Study","model":"DTrOCR","rank_in_archive_order":1,"of":7,"metrics":{"Accuracy (%)":"89.6"},"uses_additional_data":false},{"leaderboard":"/sota/optical-character-recognition-on-benchmarking","task":"Optical Character Recognition (OCR)","dataset":"Benchmarking Chinese Text Recognition: Datasets, Baselines, and an Empirical Study","model":"DTrOCR 105M","rank_in_archive_order":2,"of":7,"metrics":{"Accuracy (%)":"89.6"},"uses_additional_data":false},{"leaderboard":"/sota/scene-text-recognition-on-cute80","task":"Scene Text Recognition","dataset":"CUTE80","model":"DTrOCR 105M","rank_in_archive_order":6,"of":18,"metrics":{"Accuracy":"99.1"},"uses_additional_data":false},{"leaderboard":"/sota/scene-text-recognition-on-icdar2013","task":"Scene Text Recognition","dataset":"ICDAR2013","model":"DTrOCR 105M","rank_in_archive_order":2,"of":38,"metrics":{"Accuracy":"99.4"},"uses_additional_data":false},{"leaderboard":"/sota/scene-text-recognition-on-icdar2015","task":"Scene Text Recognition","dataset":"ICDAR2015","model":"DTrOCR 105M","rank_in_archive_order":1,"of":27,"metrics":{"Accuracy":"93.5"},"uses_additional_data":false},{"leaderboard":"/sota/scene-text-recognition-on-iiit5k","task":"Scene Text Recognition","dataset":"IIIT5k","model":"DTrOCR 105M","rank_in_archive_order":2,"of":17,"metrics":{"Accuracy":"99.6"},"uses_additional_data":false},{"leaderboard":"/sota/scene-text-recognition-on-svt","task":"Scene Text Recognition","dataset":"SVT","model":"DTrOCR 105M","rank_in_archive_order":2,"of":37,"metrics":{"Accuracy":"98.9"},"uses_additional_data":false},{"leaderboard":"/sota/scene-text-recognition-on-svtp","task":"Scene Text Recognition","dataset":"SVTP","model":"DTrOCR 105M","rank_in_archive_order":1,"of":17,"metrics":{"Accuracy":"98.6"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2308.15996","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}