{"url":"/sota/action-recognition-in-videos-on-ntu-rgbd-120","task":{"name":"Action Recognition","url":"/task/action-recognition-in-videos","note":null},"dataset":{"name":"NTU RGB+D 120","url":"/dataset/ntu-rgb-d-120"},"category":"Computer Vision","categories":["Computer Vision","Time Series"],"category_note":null,"description":"**Action Recognition** is a computer vision task that involves recognizing human actions in videos or images. The goal is to classify and categorize the actions being performed in the video or image into a predefined set of action classes.\r\n\r\nIn the video domain, it is an open question whether training an action classification network on a sufficiently large dataset, will give a similar boost in performance when applied to a different temporal task or dataset. The challenges of building video datasets has meant that most popular benchmarks for action recognition are small, having on the order of 10k videos. \r\n\r\nPlease note some benchmarks may be located in the [Action Classification](https://paperswithcode.com/task/action-classification) or [Video Classification](https://paperswithcode.com/task/video-classification) tasks, e.g. Kinetics-400.","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 (Cross-Setup)","Accuracy (Cross-Subject)"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Accuracy (Cross-Setup)":"higher","Accuracy (Cross-Subject)":"higher"}},"counts":{"rows":21,"rows_with_code":17,"rows_with_paper_page":21,"rows_dated":21,"rows_using_additional_data":5},"rows":[{"rank_in_archive_order":1,"model":"DSCNet (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"96.7","Accuracy (Cross-Subject)":"95.6"},"uses_additional_data":false,"paper_date":"2023-12-28","paper":"/paper/a-dense-sparse-complementary-network-for","paper_url":"https://www.sciencedirect.com/science/article/pii/S0957417423035637","paper_title":"A Dense-Sparse Complementary Network for Human Action Recognition based on RGB and Skeleton Modalities","code":"https://github.com/Maxchengqin/DSCNet","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"PoseC3D (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"96.4","Accuracy (Cross-Subject)":"95.3"},"uses_additional_data":false,"paper_date":"2021-04-28","paper":"/paper/revisiting-skeleton-based-action-recognition","paper_url":"https://arxiv.org/abs/2104.13586v2","paper_title":"Revisiting Skeleton-based Action Recognition","code":"https://github.com/open-mmlab/mmaction2","n_code_links":4,"syntology":null},{"rank_in_archive_order":3,"model":"π-ViT (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"96.1","Accuracy (Cross-Subject)":"95.1"},"uses_additional_data":true,"paper_date":"2023-11-30","paper":"/paper/just-add-p-pose-induced-video-transformers","paper_url":"https://arxiv.org/abs/2311.18840v1","paper_title":"Just Add $π$! Pose Induced Video Transformers for Understanding Activities of Daily Living","code":"https://github.com/dominickrei/pi-vit","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"MMNet (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"94.4","Accuracy (Cross-Subject)":"92.9"},"uses_additional_data":true,"paper_date":"2022-05-26","paper":"/paper/mmnet-a-model-based-multimodal-network-for","paper_url":"https://ieeexplore.ieee.org/abstract/document/9782511","paper_title":"MMNet: A Model-Based Multimodal Network for Human Action Recognition in RGB-D Videos","code":"https://github.com/bruceyo/MMNet","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"EPP-Net (Parsing + Pose)","metrics":{"Accuracy (Cross-Setup)":"92.8","Accuracy (Cross-Subject)":"91.1"},"uses_additional_data":false,"paper_date":"2024-01-04","paper":"/paper/explore-human-parsing-modality-for-action-1","paper_url":"https://arxiv.org/abs/2401.02138v1","paper_title":"Explore Human Parsing Modality for Action Recognition","code":"https://github.com/liujf69/EPP-Net-Action","n_code_links":1,"syntology":{"n_ran":7,"n_unverified":3,"n_samples":10,"n_pointer_only_licence":10}},{"rank_in_archive_order":6,"model":"STAR-Transformer (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"92.7","Accuracy (Cross-Subject)":"90.3"},"uses_additional_data":false,"paper_date":"2022-10-14","paper":"/paper/star-transformer-a-spatio-temporal-cross-1","paper_url":"https://arxiv.org/abs/2210.07503v1","paper_title":"STAR-Transformer: A Spatio-temporal Cross Attention Transformer for Human Action Recognition","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"EPAM-Net","metrics":{"Accuracy (Cross-Setup)":"92.4","Accuracy (Cross-Subject)":"94.3"},"uses_additional_data":false,"paper_date":"2024-08-10","paper":"/paper/epam-net-an-efficient-pose-driven-attention","paper_url":"https://arxiv.org/abs/2408.05421v1","paper_title":"EPAM-Net: An Efficient Pose-driven Attention-guided Multimodal Network for Video Action Recognition","code":"https://github.com/ahmed-nady/multimodal-action-recognition","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"π-ViT (RGB only)","metrics":{"Accuracy (Cross-Setup)":"91.9","Accuracy (Cross-Subject)":"92.9"},"uses_additional_data":true,"paper_date":"2023-11-30","paper":"/paper/just-add-p-pose-induced-video-transformers","paper_url":"https://arxiv.org/abs/2311.18840v1","paper_title":"Just Add $π$! Pose Induced Video Transformers for Understanding Activities of Daily Living","code":"https://github.com/dominickrei/pi-vit","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"IPP-Net (Parsing + Pose)","metrics":{"Accuracy (Cross-Setup)":"91.7","Accuracy (Cross-Subject)":"90.0"},"uses_additional_data":false,"paper_date":"2023-07-16","paper":"/paper/integrating-human-parsing-and-pose-network","paper_url":"https://arxiv.org/abs/2307.07977v1","paper_title":"Integrating Human Parsing and Pose Network for Human Action Recognition","code":"https://github.com/liujf69/ipp-net-parsing","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"3DA (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"91.4","Accuracy (Cross-Subject)":"90.5"},"uses_additional_data":false,"paper_date":"2022-12-12","paper":"/paper/cross-modal-learning-with-3d-deformable","paper_url":"https://arxiv.org/abs/2212.05638v3","paper_title":"Cross-Modal Learning with 3D Deformable Attention for Action Recognition","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":11,"model":"JPFormer(Pose)","metrics":{"Accuracy (Cross-Setup)":"91.4","Accuracy (Cross-Subject)":"89.4"},"uses_additional_data":false,"paper_date":"2024-07-30","paper":"/paper/joint-partition-group-attention-for-skeleton","paper_url":"https://www.sciencedirect.com/science/article/pii/S0165168424002111","paper_title":"Joint-Partition Group Attention for skeleton-based action recognition","code":"https://github.com/HuCui2022/JPFormer","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"DSTSA-GCN","metrics":{"Accuracy (Cross-Setup)":"90.97","Accuracy (Cross-Subject)":"89.12"},"uses_additional_data":false,"paper_date":"2025-01-21","paper":"/paper/dstsa-gcn-advancing-skeleton-based-gesture","paper_url":"https://arxiv.org/abs/2501.12086v1","paper_title":"DSTSA-GCN: Advancing Skeleton-Based Gesture Recognition with Semantic-Aware Spatio-Temporal Topology Modeling","code":"https://github.com/HuCui2022/DSTSA-GCN_Gesture","n_code_links":1,"syntology":null},{"rank_in_archive_order":13,"model":"VPN++ (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"90.7","Accuracy (Cross-Subject)":"92.5"},"uses_additional_data":true,"paper_date":"2021-05-17","paper":"/paper/vpn-rethinking-video-pose-embeddings-for","paper_url":"https://arxiv.org/abs/2105.08141v1","paper_title":"VPN++: Rethinking Video-Pose embeddings for understanding Activities of Daily Living","code":"https://github.com/srijandas07/vpnplusplus","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":14,"model":"DVANet (RGB only)","metrics":{"Accuracy (Cross-Setup)":"90.4","Accuracy (Cross-Subject)":"91.6"},"uses_additional_data":false,"paper_date":"2023-12-10","paper":"/paper/dvanet-disentangling-view-and-action-features","paper_url":"https://arxiv.org/abs/2312.05719v1","paper_title":"DVANet: Disentangling View and Action Features for Multi-View Action Recognition","code":"https://github.com/NyleSiddiqui/MultiView_Actions","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"ViewCon (RGB)","metrics":{"Accuracy (Cross-Setup)":"87.5","Accuracy (Cross-Subject)":"85.6"},"uses_additional_data":false,"paper_date":"2023-01-03","paper":"/paper/multi-view-action-recognition-using","paper_url":"https://openaccess.thecvf.com/content/WACV2023/html/Shah_Multi-View_Action_Recognition_Using_Contrastive_Learning_WACV_2023_paper.html","paper_title":"Multi-View Action Recognition Using Contrastive Learning","code":"https://github.com/kshah33/viewcon","n_code_links":1,"syntology":null},{"rank_in_archive_order":16,"model":"VPN (RGB + Pose)","metrics":{"Accuracy (Cross-Setup)":"86.3","Accuracy (Cross-Subject)":"87.8"},"uses_additional_data":true,"paper_date":"2020-07-06","paper":"/paper/vpn-learning-video-pose-embedding-for","paper_url":"https://arxiv.org/abs/2007.03056v1","paper_title":"VPN: Learning Video-Pose Embedding for Activities of Daily Living","code":"https://github.com/srijandas07/VPN","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"ST-GCN + AS-GCN w/DH-TCN","metrics":{"Accuracy (Cross-Setup)":"78.3","Accuracy (Cross-Subject)":"79.2"},"uses_additional_data":false,"paper_date":"2019-12-20","paper":"/paper/vertex-feature-encoding-and-hierarchical","paper_url":"https://arxiv.org/abs/1912.09745v1","paper_title":"Vertex Feature Encoding and Hierarchical Temporal Modeling in a Spatial-Temporal Graph Convolutional Network for Action Recognition","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":18,"model":"Gimme Signals (AIS)","metrics":{"Accuracy (Cross-Setup)":"70.8","Accuracy (Cross-Subject)":"71.59"},"uses_additional_data":false,"paper_date":"2020-03-13","paper":"/paper/gimme-signals-discriminative-signal-encoding","paper_url":"https://arxiv.org/abs/2003.06156v2","paper_title":"Gimme Signals: Discriminative signal encoding for multimodal activity recognition","code":"https://github.com/raphaelmemmesheimer/gimme_signals_action_recognition","n_code_links":2,"syntology":null},{"rank_in_archive_order":19,"model":"TSRJI","metrics":{"Accuracy (Cross-Setup)":"67.9","Accuracy (Cross-Subject)":"62.8"},"uses_additional_data":false,"paper_date":"2019-09-11","paper":"/paper/skeleton-image-representation-for-3d-action","paper_url":"https://arxiv.org/abs/1909.05704v1","paper_title":"Skeleton Image Representation for 3D Action Recognition based on Tree Structure and Reference Joints","code":"https://github.com/carloscaetano/skeleton-images","n_code_links":1,"syntology":null},{"rank_in_archive_order":20,"model":"Skelemotion + Yang et al. (skeleton only)","metrics":{"Accuracy (Cross-Setup)":"66.9","Accuracy (Cross-Subject)":"67.7"},"uses_additional_data":false,"paper_date":"2019-07-30","paper":"/paper/skelemotion-a-new-representation-of-skeleton","paper_url":"https://arxiv.org/abs/1907.13025v1","paper_title":"SkeleMotion: A New Representation of Skeleton Joint Sequences Based on Motion Information for 3D Action Recognition","code":"https://github.com/carloscaetano/skeleton-images","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":21,"model":"Body Pose Evolution Map","metrics":{"Accuracy (Cross-Setup)":"64.6","Accuracy (Cross-Subject)":"66.9"},"uses_additional_data":false,"paper_date":"2018-06-01","paper":"/paper/recognizing-human-actions-as-the-evolution-of","paper_url":"http://openaccess.thecvf.com/content_cvpr_2018/html/Liu_Recognizing_Human_Actions_CVPR_2018_paper.html","paper_title":"Recognizing Human Actions as the Evolution of Pose Estimation Maps","code":null,"n_code_links":0,"syntology":null}],"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 6,885 of the 9,623 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":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+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":3,"rows_with_any_sample_ran":3,"distinct_papers_with_graph_line":3,"distinct_papers_with_any_sample_ran":3,"samples_over_distinct_papers":{"n_ran":11,"n_unverified":3,"n_samples":14,"n_pointer_only_licence":14,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":11,"n_unverified":3,"n_samples":14,"n_pointer_only_licence":14,"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"}}}