{"url":"/sota/document-image-classification-on-rvl-cdip","task":{"name":"Document Image Classification","url":"/task/document-image-classification","note":null},"dataset":{"name":"RVL-CDIP","url":"/dataset/rvl-cdip"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"Document image classification is the task of classifying documents based on images of their contents.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [Real-Time Document Image Classification using Deep CNN and Extreme Learning Machines](https://arxiv.org/pdf/1711.05862v1.pdf) )</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Accuracy","Parameters"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Accuracy":"higher","Parameters":null}},"counts":{"rows":31,"rows_with_code":25,"rows_with_paper_page":31,"rows_dated":31,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"EAML","metrics":{"Accuracy":"97.70%"},"uses_additional_data":false,"paper_date":"2023-05-11","paper":"/paper/eaml-ensemble-self-attention-based-mutual","paper_url":"https://arxiv.org/abs/2305.06923v1","paper_title":"EAML: Ensemble Self-Attention-based Mutual Learning Network for Document Image Classification","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"Cross-Modal","metrics":{"Accuracy":"97.05%","Parameters":"197M"},"uses_additional_data":false,"paper_date":"2020-06-16","paper":"/paper/visual-and-textual-deep-feature-fusion-for","paper_url":"https://openaccess.thecvf.com/content_CVPRW_2020/html/w34/Bakkali_Visual_and_Textual_Deep_Feature_Fusion_for_Document_Image_Classification_CVPRW_2020_paper.html","paper_title":"Visual and Textual Deep Feature Fusion for Document Image Classification","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"DocFormerBASE","metrics":{"Accuracy":"96.17%","Parameters":"183M"},"uses_additional_data":false,"paper_date":"2021-06-22","paper":"/paper/docformer-end-to-end-transformer-for-document","paper_url":"https://arxiv.org/abs/2106.11539v2","paper_title":"DocFormer: End-to-End Transformer for Document Understanding","code":"https://github.com/shabie/docformer","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":4,"model":"LayoutLMV3Large","metrics":{"Accuracy":"95.93%","Parameters":"368M"},"uses_additional_data":false,"paper_date":"2022-04-18","paper":"/paper/layoutlmv3-pre-training-for-document-ai-with","paper_url":"https://arxiv.org/abs/2204.08387v3","paper_title":"LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking","code":"https://github.com/huggingface/transformers","n_code_links":4,"syntology":null},{"rank_in_archive_order":5,"model":"LiLT[EN-R]BASE","metrics":{"Accuracy":"95.68%"},"uses_additional_data":false,"paper_date":"2022-02-28","paper":"/paper/lilt-a-simple-yet-effective-language","paper_url":"https://arxiv.org/abs/2202.13669v1","paper_title":"LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding","code":"https://github.com/huggingface/transformers","n_code_links":5,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"LayoutLMv2LARGE","metrics":{"Accuracy":"95.64%"},"uses_additional_data":false,"paper_date":"2020-12-29","paper":"/paper/layoutlmv2-multi-modal-pre-training-for","paper_url":"https://arxiv.org/abs/2012.14740v4","paper_title":"LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding","code":"https://github.com/huggingface/transformers","n_code_links":9,"syntology":null},{"rank_in_archive_order":7,"model":"TILT-Large","metrics":{"Accuracy":"95.52%"},"uses_additional_data":false,"paper_date":"2021-02-18","paper":"/paper/going-full-tilt-boogie-on-document","paper_url":"https://arxiv.org/abs/2102.09550v3","paper_title":"Going Full-TILT Boogie on Document Understanding with Text-Image-Layout Transformer","code":"https://github.com/uakarsh/TiLT-Implementation","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"DocFormer large","metrics":{"Accuracy":"95.50%","Parameters":"536M"},"uses_additional_data":false,"paper_date":"2021-06-22","paper":"/paper/docformer-end-to-end-transformer-for-document","paper_url":"https://arxiv.org/abs/2106.11539v2","paper_title":"DocFormer: End-to-End Transformer for Document Understanding","code":"https://github.com/shabie/docformer","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":9,"model":"LayoutLMv3BASE","metrics":{"Accuracy":"95.44%","Parameters":"133M"},"uses_additional_data":false,"paper_date":"2022-04-18","paper":"/paper/layoutlmv3-pre-training-for-document-ai-with","paper_url":"https://arxiv.org/abs/2204.08387v3","paper_title":"LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking","code":"https://github.com/huggingface/transformers","n_code_links":4,"syntology":null},{"rank_in_archive_order":10,"model":"Donut","metrics":{"Accuracy":"95.3%"},"uses_additional_data":false,"paper_date":"2021-11-30","paper":"/paper/donut-document-understanding-transformer","paper_url":"https://arxiv.org/abs/2111.15664v5","paper_title":"OCR-free Document Understanding Transformer","code":"https://github.com/clovaai/donut","n_code_links":5,"syntology":{"n_ran":0,"n_unverified":9,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"TILT-Base","metrics":{"Accuracy":"95.25%"},"uses_additional_data":false,"paper_date":"2021-02-18","paper":"/paper/going-full-tilt-boogie-on-document","paper_url":"https://arxiv.org/abs/2102.09550v3","paper_title":"Going Full-TILT Boogie on Document Understanding with Text-Image-Layout Transformer","code":"https://github.com/uakarsh/TiLT-Implementation","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"LayoutLMv2BASE","metrics":{"Accuracy":"95.25%","Parameters":"200M"},"uses_additional_data":false,"paper_date":"2020-12-29","paper":"/paper/layoutlmv2-multi-modal-pre-training-for","paper_url":"https://arxiv.org/abs/2012.14740v4","paper_title":"LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding","code":"https://github.com/huggingface/transformers","n_code_links":9,"syntology":null},{"rank_in_archive_order":13,"model":"LayoutXLM","metrics":{"Accuracy":"95.21%"},"uses_additional_data":false,"paper_date":"2021-04-18","paper":"/paper/layoutxlm-multimodal-pre-training-for","paper_url":"https://arxiv.org/abs/2104.08836v3","paper_title":"LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding","code":"https://github.com/huggingface/transformers","n_code_links":6,"syntology":null},{"rank_in_archive_order":14,"model":"StrucTexTv2 (large)","metrics":{"Accuracy":"94.62%","Parameters":"238M"},"uses_additional_data":false,"paper_date":"2023-03-01","paper":"/paper/structextv2-masked-visual-textual-prediction","paper_url":"https://arxiv.org/abs/2303.00289v1","paper_title":"StrucTexTv2: Masked Visual-Textual Prediction for Document Image Pre-training","code":"https://github.com/PaddlePaddle/VIMER/tree/main/StrucTexT/v2","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"Pre-trained LayoutLM","metrics":{"Accuracy":"94.42%","Parameters":"160M"},"uses_additional_data":false,"paper_date":"2019-12-31","paper":"/paper/layoutlm-pre-training-of-text-and-layout-for","paper_url":"https://arxiv.org/abs/1912.13318v5","paper_title":"LayoutLM: Pre-training of Text and Layout for Document Image Understanding","code":"https://github.com/huggingface/transformers","n_code_links":19,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":1}},{"rank_in_archive_order":16,"model":"DoPTA","metrics":{"Accuracy":"94.12%","Parameters":"85M"},"uses_additional_data":false,"paper_date":"2024-12-17","paper":"/paper/dopta-improving-document-layout-analysis","paper_url":"https://arxiv.org/abs/2412.12902v1","paper_title":"DoPTA: Improving Document Layout Analysis using Patch-Text Alignment","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":17,"model":"DocXClassifier-B","metrics":{"Accuracy":"94.00%","Parameters":"95.4M"},"uses_additional_data":false,"paper_date":"2022-03-17","paper":"/paper/docxclassifier-high-performance-explainable","paper_url":"https://www.techrxiv.org/articles/preprint/DocXClassifier_High_Performance_Explainable_Deep_Network_for_Document_Image_Classification/19310489","paper_title":"DocXClassifier: High Performance Explainable Deep Network for Document Image Classification","code":"https://github.com/saifullah3396/docxclassifier","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"StrucTexTv2 (small)","metrics":{"Accuracy":"93.4%","Parameters":"28M"},"uses_additional_data":false,"paper_date":"2023-03-01","paper":"/paper/structextv2-masked-visual-textual-prediction","paper_url":"https://arxiv.org/abs/2303.00289v1","paper_title":"StrucTexTv2: Masked Visual-Textual Prediction for Document Image Pre-training","code":"https://github.com/PaddlePaddle/VIMER/tree/main/StrucTexT/v2","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"VLCDoC","metrics":{"Accuracy":"93.19%","Parameters":"217M"},"uses_additional_data":false,"paper_date":"2022-05-24","paper":"/paper/vlcdoc-vision-language-contrastive-pre","paper_url":"https://arxiv.org/abs/2205.12029v3","paper_title":"VLCDoC: Vision-Language Contrastive Pre-Training Model for Cross-Modal Document Classification","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":20,"model":"TransferDoc","metrics":{"Accuracy":"93.18%","Parameters":"221M"},"uses_additional_data":false,"paper_date":"2023-09-11","paper":"/paper/transferdoc-a-self-supervised-transferable","paper_url":"https://arxiv.org/abs/2309.05756v3","paper_title":"GlobalDoc: A Cross-Modal Vision-Language Framework for Real-World Document Image Retrieval and Classification","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":21,"model":"Multimodal (ResNet50)","metrics":{"Accuracy":"92.7%","Parameters":"57M"},"uses_additional_data":false,"paper_date":"2023-01-16","paper":"/paper/multimodal-side-tuning-for-document","paper_url":"https://arxiv.org/abs/2301.07502v2","paper_title":"Multimodal Side-Tuning for Document Classification","code":"https://github.com/thezingaro/multimodal-side-tuning","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"DiT-L","metrics":{"Accuracy":"92.69%","Parameters":"304M"},"uses_additional_data":false,"paper_date":"2022-03-04","paper":"/paper/dit-self-supervised-pre-training-for-document","paper_url":"https://arxiv.org/abs/2203.02378v3","paper_title":"DiT: Self-supervised Pre-training for Document Image Transformer","code":"https://github.com/huggingface/transformers","n_code_links":4,"syntology":{"n_ran":0,"n_unverified":11,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"Pre-trained EfficientNet","metrics":{"Accuracy":"92.31%"},"uses_additional_data":false,"paper_date":"2020-06-16","paper":"/paper/improving-accuracy-and-speeding-up-document","paper_url":"https://arxiv.org/abs/2006.09141v1","paper_title":"Improving accuracy and speeding up Document Image Classification through parallel systems","code":"https://github.com/javiferran/document-classification","n_code_links":1,"syntology":null},{"rank_in_archive_order":24,"model":"Transfer Learning from VGG16 trained on Imagenet","metrics":{"Accuracy":"92.21%"},"uses_additional_data":false,"paper_date":"2018-01-29","paper":"/paper/document-image-classification-with-intra","paper_url":"http://arxiv.org/abs/1801.09321v3","paper_title":"Document Image Classification with Intra-Domain Transfer Learning and Stacked Generalization of Deep Convolutional Neural Networks","code":"https://github.com/microsoft/unilm/tree/master/layoutlm","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":25,"model":"Multimodal (MobileNetV2)","metrics":{"Accuracy":"92.2%","Parameters":"12M"},"uses_additional_data":false,"paper_date":"2023-01-16","paper":"/paper/multimodal-side-tuning-for-document","paper_url":"https://arxiv.org/abs/2301.07502v2","paper_title":"Multimodal Side-Tuning for Document Classification","code":"https://github.com/thezingaro/multimodal-side-tuning","n_code_links":1,"syntology":null},{"rank_in_archive_order":26,"model":"DiT-B","metrics":{"Accuracy":"92.11%","Parameters":"87M"},"uses_additional_data":false,"paper_date":"2022-03-04","paper":"/paper/dit-self-supervised-pre-training-for-document","paper_url":"https://arxiv.org/abs/2203.02378v3","paper_title":"DiT: Self-supervised Pre-training for Document Image Transformer","code":"https://github.com/huggingface/transformers","n_code_links":4,"syntology":{"n_ran":0,"n_unverified":11,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":27,"model":"BEiT-B","metrics":{"Accuracy":"91.09%","Parameters":"87M"},"uses_additional_data":false,"paper_date":"2021-06-15","paper":"/paper/beit-bert-pre-training-of-image-transformers","paper_url":"https://arxiv.org/abs/2106.08254v2","paper_title":"BEiT: BERT Pre-Training of Image Transformers","code":"https://github.com/huggingface/transformers","n_code_links":14,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":28,"model":"Transfer Learning from AlexNet, VGG-16, GoogLeNet and ResNet50","metrics":{"Accuracy":"90.97%"},"uses_additional_data":false,"paper_date":"2017-04-11","paper":"/paper/cutting-the-error-by-half-investigation-of","paper_url":"http://arxiv.org/abs/1704.03557v1","paper_title":"Cutting the Error by Half: Investigation of Very Deep CNN and Advanced Training Strategies for Document Image Classification","code":"https://github.com/microsoft/unilm/tree/master/layoutlm","n_code_links":5,"syntology":null},{"rank_in_archive_order":29,"model":"AlexNet + spatial pyramidal pooling + image resizing","metrics":{"Accuracy":"90.94%"},"uses_additional_data":false,"paper_date":"2017-08-10","paper":"/paper/analysis-of-convolutional-neural-networks-for","paper_url":"http://arxiv.org/abs/1708.03273v1","paper_title":"Analysis of Convolutional Neural Networks for Document Image Classification","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":30,"model":"DeiT-B","metrics":{"Accuracy":"90.32%","Parameters":" 87M"},"uses_additional_data":false,"paper_date":"2020-12-23","paper":"/paper/training-data-efficient-image-transformers","paper_url":"https://arxiv.org/abs/2012.12877v2","paper_title":"Training data-efficient image transformers & distillation through attention","code":"https://github.com/huggingface/transformers","n_code_links":40,"syntology":{"n_ran":12,"n_unverified":7,"n_samples":19,"n_pointer_only_licence":3}},{"rank_in_archive_order":31,"model":"Roberta base","metrics":{"Accuracy":"90.06","Parameters":"125M"},"uses_additional_data":false,"paper_date":"2019-07-26","paper":"/paper/roberta-a-robustly-optimized-bert-pretraining","paper_url":"https://arxiv.org/abs/1907.11692v1","paper_title":"RoBERTa: A Robustly Optimized BERT Pretraining Approach","code":"https://github.com/huggingface/transformers","n_code_links":67,"syntology":{"n_ran":22,"n_unverified":26,"n_samples":48,"n_pointer_only_licence":23}}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":11,"rows_with_any_sample_ran":8,"distinct_papers_with_graph_line":9,"distinct_papers_with_any_sample_ran":7,"samples_over_distinct_papers":{"n_ran":50,"n_unverified":65,"n_samples":115,"n_pointer_only_licence":27,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":55,"n_unverified":81,"n_samples":136,"n_pointer_only_licence":27,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}