{"url":"/sota/human-pose-forecasting-on-human36m","task":{"name":"Human Pose Forecasting","url":"/task/human-pose-forecasting","note":null},"dataset":{"name":"Human3.6M","url":"/dataset/human3-6m"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"Human pose forecasting is the task of detecting and predicting future human poses.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [EgoPose](https://github.com/Khrylx/EgoPose) )</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":["Average MPJPE (mm) @ 400ms","Average MPJPE (mm) @ 1000 ms","MAR, walking, 400ms","MAR, walking, 1,000ms","APD","ADE","FDE","MMADE","MMFDE","CMD","FID","Average NDMS at 4s "],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Average MPJPE (mm) @ 400ms":null,"Average MPJPE (mm) @ 1000 ms":null,"MAR, walking, 400ms":null,"MAR, walking, 1,000ms":null,"APD":null,"ADE":null,"FDE":null,"MMADE":null,"MMFDE":null,"CMD":null,"FID":"lower","Average NDMS at 4s ":null}},"counts":{"rows":33,"rows_with_code":24,"rows_with_paper_page":33,"rows_dated":33,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"GCNext","metrics":{"Average MPJPE (mm) @ 1000 ms":"64.7","Average MPJPE (mm) @ 400ms":"30.5"},"uses_additional_data":false,"paper_date":"2023-12-19","paper":"/paper/gcnext-towards-the-unity-of-graph","paper_url":"https://arxiv.org/abs/2312.11850v1","paper_title":"GCNext: Towards the Unity of Graph Convolutions for Human Motion Prediction","code":"https://github.com/bradleywang0416/gcnext","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"MGCN","metrics":{"Average MPJPE (mm) @ 1000 ms":"72.9","Average MPJPE (mm) @ 400ms":"34.4"},"uses_additional_data":false,"paper_date":"2023-04-11","paper":"/paper/multi-graph-convolution-network-for-pose","paper_url":"https://arxiv.org/abs/2304.04956v1","paper_title":"Multi-Graph Convolution Network for Pose Forecasting","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"FMS-AM","metrics":{"Average MPJPE (mm) @ 400ms":"34.8"},"uses_additional_data":false,"paper_date":"2023-05-19","paper":"/paper/remembering-what-is-important-a-factorised","paper_url":"https://arxiv.org/abs/2305.11394v1","paper_title":"Remembering What Is Important: A Factorised Multi-Head Retrieval and Auxiliary Memory Stabilisation Scheme for Human Motion Prediction","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"AuxFormer","metrics":{"Average MPJPE (mm) @ 1000 ms":"107","Average MPJPE (mm) @ 400ms":"54.1"},"uses_additional_data":false,"paper_date":"2023-08-17","paper":"/paper/auxiliary-tasks-benefit-3d-skeleton-based","paper_url":"https://arxiv.org/abs/2308.08942v1","paper_title":"Auxiliary Tasks Benefit 3D Skeleton-based Human Motion Prediction","code":"https://github.com/mediabrain-sjtu/auxformer","n_code_links":1,"syntology":{"n_ran":13,"n_unverified":6,"n_samples":19,"n_pointer_only_licence":19}},{"rank_in_archive_order":5,"model":"EqMotion","metrics":{"Average MPJPE (mm) @ 1000 ms":"106.9","Average MPJPE (mm) @ 400ms":"55.0"},"uses_additional_data":false,"paper_date":"2023-03-20","paper":"/paper/eqmotion-equivariant-multi-agent-motion","paper_url":"https://arxiv.org/abs/2303.10876v2","paper_title":"EqMotion: Equivariant Multi-agent Motion Prediction with Invariant Interaction Reasoning","code":"https://github.com/mediabrain-sjtu/eqmotion","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":1,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"Sun et al.","metrics":{"Average MPJPE (mm) @ 1000 ms":"109.2","Average MPJPE (mm) @ 400ms":"55.5"},"uses_additional_data":false,"paper_date":"2023-05-08","paper":"/paper/towards-accurate-human-motion-prediction-via","paper_url":"https://arxiv.org/abs/2305.04443v1","paper_title":"Towards Accurate Human Motion Prediction via Iterative Refinement","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"ResChunk","metrics":{"Average MPJPE (mm) @ 1000 ms":"81.95","Average MPJPE (mm) @ 400ms":"56.41"},"uses_additional_data":false,"paper_date":"2023-08-31","paper":"/paper/multiscale-residual-learning-of-graph","paper_url":"https://arxiv.org/abs/2308.16801v1","paper_title":"Multiscale Residual Learning of Graph Convolutional Sequence Chunks for Human Motion Prediction","code":"https://github.com/mohsenzand/reschunk","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"GraphMixer","metrics":{"Average MPJPE (mm) @ 1000 ms":"108.6","Average MPJPE (mm) @ 400ms":"56.7"},"uses_additional_data":false,"paper_date":"2023-04-07","paper":"/paper/a-mixer-layer-is-worth-one-graph-convolution","paper_url":"https://arxiv.org/abs/2304.03532v2","paper_title":"Graph-Guided MLP-Mixer for Skeleton-Based Human Motion Prediction","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":9,"model":"siMLPe","metrics":{"Average MPJPE (mm) @ 1000 ms":"109.4","Average MPJPE (mm) @ 400ms":"57.3"},"uses_additional_data":false,"paper_date":"2022-07-04","paper":"/paper/back-to-mlp-a-simple-baseline-for-human","paper_url":"https://arxiv.org/abs/2207.01567v3","paper_title":"Back to MLP: A Simple Baseline for Human Motion Prediction","code":"https://github.com/dulucas/simlpe","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":4,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"PGBIG","metrics":{"Average MPJPE (mm) @ 1000 ms":"110.3","Average MPJPE (mm) @ 400ms":"58.5","MAR, walking, 1,000ms":"0.69","MAR, walking, 400ms":"0.54"},"uses_additional_data":false,"paper_date":"2022-03-30","paper":"/paper/progressively-generating-better-initial","paper_url":"https://arxiv.org/abs/2203.16051v1","paper_title":"Progressively Generating Better Initial Guesses Towards Next Stages for High-Quality Human Motion Prediction","code":"https://github.com/705062791/pgbig","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":4,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":11,"model":"MotionMixer","metrics":{"Average MPJPE (mm) @ 1000 ms":"111.0","Average MPJPE (mm) @ 400ms":"59.3","MAR, walking, 1,000ms":"0.73","MAR, walking, 400ms":"0.58"},"uses_additional_data":false,"paper_date":"2022-07-01","paper":"/paper/motionmixer-mlp-based-3d-human-body-pose","paper_url":"https://arxiv.org/abs/2207.00499v1","paper_title":"MotionMixer: MLP-based 3D Human Body Pose Forecasting","code":"https://github.com/motionmlp/motionmixer","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":1,"n_samples":5,"n_pointer_only_licence":5}},{"rank_in_archive_order":12,"model":"UNIMASK-M","metrics":{"Average MPJPE (mm) @ 1000 ms":"112.1","Average MPJPE (mm) @ 400ms":"61.6"},"uses_additional_data":false,"paper_date":"2023-08-14","paper":"/paper/a-unified-masked-autoencoder-with-patchified","paper_url":"https://arxiv.org/abs/2308.07301v2","paper_title":"A Unified Masked Autoencoder with Patchified Skeletons for Motion Synthesis","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":13,"model":"MSR-GCN","metrics":{"Average MPJPE (mm) @ 1000 ms":"114.2","Average MPJPE (mm) @ 400ms":"62.9"},"uses_additional_data":false,"paper_date":"2021-08-16","paper":"/paper/msr-gcn-multi-scale-residual-graph","paper_url":"https://arxiv.org/abs/2108.07152v2","paper_title":"MSR-GCN: Multi-Scale Residual Graph Convolution Networks for Human Motion Prediction","code":"https://github.com/Droliven/MSRGCN","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":1,"n_samples":7,"n_pointer_only_licence":7}},{"rank_in_archive_order":14,"model":"LTD-GCN","metrics":{"Average MPJPE (mm) @ 1000 ms":"113.0","Average MPJPE (mm) @ 400ms":"63.5","MAR, walking, 1,000ms":"0.67","MAR, walking, 400ms":"0.56"},"uses_additional_data":false,"paper_date":"2019-08-15","paper":"/paper/learning-trajectory-dependencies-for-human","paper_url":"https://arxiv.org/abs/1908.05436v3","paper_title":"Learning Trajectory Dependencies for Human Motion Prediction","code":"https://github.com/wei-mao-2019/LearnTrajDep","n_code_links":5,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"STS-GCN","metrics":{"Average MPJPE (mm) @ 1000 ms":"117.0","Average MPJPE (mm) @ 400ms":"65.8","MAR, walking, 1,000ms":"0.87","MAR, walking, 400ms":"0.55"},"uses_additional_data":false,"paper_date":"2021-10-09","paper":"/paper/space-time-separable-graph-convolutional-1","paper_url":"https://arxiv.org/abs/2110.04573v1","paper_title":"Space-Time-Separable Graph Convolutional Network for Pose Forecasting","code":"https://github.com/fraluca/stsgcn","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"AnyPose1","metrics":{"Average MPJPE (mm) @ 1000 ms":"128.2","Average MPJPE (mm) @ 400ms":"80.6"},"uses_additional_data":false,"paper_date":"2023-09-09","paper":"/paper/anypose-anytime-3d-human-pose-forecasting-via","paper_url":"https://arxiv.org/abs/2309.04840v1","paper_title":"AnyPose: Anytime 3D Human Pose Forecasting via Neural Ordinary Differential Equations","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":17,"model":"Forecast LSTM","metrics":{"APD":"3070","Average MPJPE (mm) @ 1000 ms":"149.2","Average MPJPE (mm) @ 400ms":"80.8","Average NDMS at 4s ":"0.465","MAR, walking, 400ms":"0.91"},"uses_additional_data":false,"paper_date":"2021-12-01","paper":"/paper/pages-iai-uni-bonn-de-https-pages-iai-uni","paper_url":"https://pages.iai.uni-bonn.de/gall_juergen/download/jgall_forecastintention_3dv21.pdf","paper_title":"Intention-based Long-Term Human Motion Anticipation","code":"https://github.com/zaverichintan/Intention-based-Long-Term-Human-Motion-Anticipation","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"BC+WGAIL-div","metrics":{"MAR, walking, 1,000ms":"0.69","MAR, walking, 400ms":"0.59"},"uses_additional_data":false,"paper_date":"2019-09-08","paper":"/paper/imitation-learning-for-human-pose-prediction","paper_url":"https://arxiv.org/abs/1909.03449v1","paper_title":"Imitation Learning for Human Pose Prediction","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":19,"model":"TP-RNN","metrics":{"MAR, walking, 1,000ms":"0.77","MAR, walking, 400ms":"0.65"},"uses_additional_data":false,"paper_date":"2018-10-23","paper":"/paper/action-agnostic-human-pose-forecasting","paper_url":"http://arxiv.org/abs/1810.09676v1","paper_title":"Action-Agnostic Human Pose Forecasting","code":"https://github.com/eddyhkchiu/pose_forecast_wacv","n_code_links":1,"syntology":null},{"rank_in_archive_order":20,"model":"SRNN","metrics":{"MAR, walking, 1,000ms":"2.13","MAR, walking, 400ms":"1.30"},"uses_additional_data":false,"paper_date":"2015-11-17","paper":"/paper/structural-rnn-deep-learning-on-spatio","paper_url":"http://arxiv.org/abs/1511.05298v3","paper_title":"Structural-RNN: Deep Learning on Spatio-Temporal Graphs","code":"https://github.com/asheshjain399/RNNexp","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"ERD","metrics":{"MAR, walking, 1,000ms":"2.38","MAR, walking, 400ms":"1.78"},"uses_additional_data":false,"paper_date":"2015-08-02","paper":"/paper/recurrent-network-models-for-human-dynamics","paper_url":"http://arxiv.org/abs/1508.00271v2","paper_title":"Recurrent Network Models for Human Dynamics","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":22,"model":"TCD","metrics":{"ADE":"356","APD":"19466","FDE":"396","MMADE":"463","MMFDE":"445"},"uses_additional_data":false,"paper_date":"2022-10-11","paper":"/paper/a-generic-diffusion-based-approach-for-3d","paper_url":"https://arxiv.org/abs/2210.05669v2","paper_title":"A generic diffusion-based approach for 3D human pose prediction in the wild","code":"https://github.com/vita-epfl/deposit","n_code_links":1,"syntology":null},{"rank_in_archive_order":23,"model":"STARS","metrics":{"ADE":"358","APD":"15884","FDE":"445","MMADE":"442","MMFDE":"471"},"uses_additional_data":false,"paper_date":"2023-02-09","paper":"/paper/diverse-human-motion-prediction-guided-by","paper_url":"https://arxiv.org/abs/2302.04860v1","paper_title":"Diverse Human Motion Prediction Guided by Multi-Level Spatial-Temporal Anchors","code":"https://github.com/sirui-xu/stars","n_code_links":1,"syntology":null},{"rank_in_archive_order":24,"model":"DiverseSampling","metrics":{"ADE":"370","APD":"15310","CMD":"11.692","FDE":"485","FID":"2.083","MMADE":"475","MMFDE":"516"},"uses_additional_data":false,"paper_date":"2022-07-15","paper":"/paper/diverse-human-motion-prediction-via-gumbel","paper_url":"https://arxiv.org/abs/2207.07351v1","paper_title":"Diverse Human Motion Prediction via Gumbel-Softmax Sampling from an Auxiliary Space","code":"https://github.com/droliven/diverse_sampling","n_code_links":2,"syntology":null},{"rank_in_archive_order":25,"model":"GSPS","metrics":{"ADE":"389","APD":"14757","CMD":"10.758","FDE":"496","FID":"2.103","MMADE":"476","MMFDE":"525"},"uses_additional_data":false,"paper_date":"2021-08-19","paper":"/paper/generating-smooth-pose-sequences-for-diverse","paper_url":"https://arxiv.org/abs/2108.08422v3","paper_title":"Generating Smooth Pose Sequences for Diverse Human Motion Prediction","code":"https://github.com/wei-mao-2019/gsps","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":7,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":26,"model":"DSF","metrics":{"ADE":"493","APD":"9330","FDE":"592","MMADE":"550","MMFDE":"599"},"uses_additional_data":false,"paper_date":"2019-07-11","paper":"/paper/diverse-trajectory-forecasting-with","paper_url":"https://arxiv.org/abs/1907.04967v2","paper_title":"Diverse Trajectory Forecasting with Determinantal Point Processes","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":27,"model":"BeLFusion","metrics":{"ADE":"372","APD":"7602","CMD":"5.988","FDE":"474","FID":"0.209","MMADE":"473","MMFDE":"507"},"uses_additional_data":false,"paper_date":"2022-11-25","paper":"/paper/belfusion-latent-diffusion-for-behavior","paper_url":"https://arxiv.org/abs/2211.14304v3","paper_title":"BeLFusion: Latent Diffusion for Behavior-Driven Human Motion Prediction","code":"https://github.com/BarqueroGerman/BeLFusion","n_code_links":1,"syntology":null},{"rank_in_archive_order":28,"model":"HP-GAN","metrics":{"ADE":"858","APD":"7214","FDE":"867","MMADE":"847","MMFDE":"858"},"uses_additional_data":false,"paper_date":"2017-11-27","paper":"/paper/hp-gan-probabilistic-3d-human-motion","paper_url":"http://arxiv.org/abs/1711.09561v1","paper_title":"HP-GAN: Probabilistic 3D human motion prediction via GAN","code":"https://github.com/ebarsoum/hpgan","n_code_links":3,"syntology":null},{"rank_in_archive_order":29,"model":"GMVAE","metrics":{"ADE":"461","APD":"6769","FDE":"555","MMADE":"524","MMFDE":"566"},"uses_additional_data":false,"paper_date":"2016-11-08","paper":"/paper/deep-unsupervised-clustering-with-gaussian","paper_url":"http://arxiv.org/abs/1611.02648v2","paper_title":"Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders","code":"https://github.com/Nat-D/GMVAE","n_code_links":4,"syntology":null},{"rank_in_archive_order":30,"model":"Pose-Knows","metrics":{"ADE":"461","APD":"6723","CMD":"6.326","FDE":"560","FID":"0.538","MMADE":"522","MMFDE":"569"},"uses_additional_data":false,"paper_date":"2017-04-28","paper":"/paper/the-pose-knows-video-forecasting-by","paper_url":"http://arxiv.org/abs/1705.00053v1","paper_title":"The Pose Knows: Video Forecasting by Generating Pose Futures","code":"https://github.com/KMarino/MMD_evalcode","n_code_links":1,"syntology":null},{"rank_in_archive_order":31,"model":"BoM","metrics":{"ADE":"448","APD":"6265","FDE":"533","MMADE":"514","MMFDE":"544"},"uses_additional_data":false,"paper_date":"2018-06-20","paper":"/paper/accurate-and-diverse-sampling-of-sequences","paper_url":"http://arxiv.org/abs/1806.07772v2","paper_title":"Accurate and Diverse Sampling of Sequences based on a \"Best of Many\" Sample Objective","code":"https://github.com/apratimbhattacharyya18/CGM_BestOfMany","n_code_links":1,"syntology":null},{"rank_in_archive_order":32,"model":"TransFusion","metrics":{"ADE":"358","APD":"5975","FDE":"468","MMADE":"506","MMFDE":"539"},"uses_additional_data":false,"paper_date":"2023-07-30","paper":"/paper/transfusion-a-practical-and-effective","paper_url":"https://arxiv.org/abs/2307.16106v1","paper_title":"TransFusion: A Practical and Effective Transformer-based Diffusion Model for 3D Human Motion Prediction","code":"https://github.com/sibotian96/TransFusion","n_code_links":1,"syntology":null},{"rank_in_archive_order":33,"model":"MT-VAE","metrics":{"ADE":"457","APD":"403","FDE":"595","MMADE":"716","MMFDE":"883"},"uses_additional_data":false,"paper_date":"2018-08-14","paper":"/paper/mt-vae-learning-motion-transformations-to","paper_url":"http://arxiv.org/abs/1808.04545v1","paper_title":"MT-VAE: Learning Motion Transformations to Generate Multimodal Human Dynamics","code":"https://github.com/xcyan/eccv18_mtvae","n_code_links":1,"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":10,"rows_with_any_sample_ran":7,"distinct_papers_with_graph_line":10,"distinct_papers_with_any_sample_ran":7,"samples_over_distinct_papers":{"n_ran":31,"n_unverified":29,"n_samples":60,"n_pointer_only_licence":37,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":31,"n_unverified":29,"n_samples":60,"n_pointer_only_licence":37,"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"}}}