{"url":"/sota/scene-graph-generation-on-visual-genome","task":{"name":"Scene Graph Generation","url":"/task/scene-graph-generation","note":null},"dataset":{"name":"Visual Genome","url":"/dataset/visual-genome"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"A scene graph is a structured representation of an image, where nodes in a scene graph correspond to object bounding boxes with their object categories, and edges correspond to their pairwise relationships between objects. The task of **Scene Graph Generation** is to generate a visually-grounded scene graph that most accurately correlates with an image.\n\n\n<span class=\"description-source\">Source: [Scene Graph Generation by Iterative Message Passing ](https://arxiv.org/abs/1701.02426)</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":["Recall@50","mean Recall @20","Recall@100","Recall@20","mean Recall @100","R@100","mR@100","mR@50","zR@100","zR@20","zR@50"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Recall@50":"higher","mean Recall @20":"higher","Recall@100":"higher","Recall@20":"higher","mean Recall @100":"higher","R@100":null,"mR@100":null,"mR@50":null,"zR@100":null,"zR@20":null,"zR@50":null}},"counts":{"rows":19,"rows_with_code":19,"rows_with_paper_page":19,"rows_dated":19,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"SpeaQ (without reweighting)","metrics":{"R@100":"36.0","Recall@100":"36.0","Recall@50":"32.9","mR@100":"14.1","mR@50":"11.8","mean Recall @100":"14.1"},"uses_additional_data":false,"paper_date":"2024-03-26","paper":"/paper/groupwise-query-specialization-and-quality","paper_url":"https://arxiv.org/abs/2403.17709v1","paper_title":"Groupwise Query Specialization and Quality-Aware Multi-Assignment for Transformer-based Visual Relationship Detection","code":"https://github.com/mlvlab/speaq","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":2,"model":"SpeaQ (with reweighting)","metrics":{"R@100":"35.5","Recall@100":"35.5","Recall@50":"32.1","mR@100":"17.6","mR@50":"15.1","mean Recall @100":"17.6"},"uses_additional_data":false,"paper_date":"2024-03-26","paper":"/paper/groupwise-query-specialization-and-quality","paper_url":"https://arxiv.org/abs/2403.17709v1","paper_title":"Groupwise Query Specialization and Quality-Aware Multi-Assignment for Transformer-based Visual Relationship Detection","code":"https://github.com/mlvlab/speaq","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":3,"model":"Causal-TDE","metrics":{"Recall@50":"31.93","mean Recall @20":"6.9"},"uses_additional_data":false,"paper_date":"2020-02-27","paper":"/paper/unbiased-scene-graph-generation-from-biased","paper_url":"https://arxiv.org/abs/2002.11949v3","paper_title":"Unbiased Scene Graph Generation from Biased Training","code":"https://github.com/KaihuaTang/Scene-Graph-Benchmark.pytorch","n_code_links":6,"syntology":{"n_ran":3,"n_unverified":3,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":4,"model":"SG-EBM","metrics":{"Recall@50":"31.74","mean Recall @20":"7.1"},"uses_additional_data":false,"paper_date":"2021-03-03","paper":"/paper/energy-based-learning-for-scene-graph","paper_url":"https://arxiv.org/abs/2103.02221v1","paper_title":"Energy-Based Learning for Scene Graph Generation","code":"https://github.com/mods333/energy-based-scene-graph","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"GPS-Net","metrics":{"Recall@50":"28.9"},"uses_additional_data":false,"paper_date":"2020-03-29","paper":"/paper/gps-net-graph-property-sensing-network-for","paper_url":"https://arxiv.org/abs/2003.12962v1","paper_title":"GPS-Net: Graph Property Sensing Network for Scene Graph Generation","code":"https://github.com/taksau/GPS-Net","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":6,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"LOGIN","metrics":{"Recall@100":"31.4","Recall@20":"22.2","Recall@50":"28.2"},"uses_additional_data":false,"paper_date":"2021-06-16","paper":"/paper/tackling-the-challenges-in-scene-graph","paper_url":"https://arxiv.org/abs/2106.08543v3","paper_title":"Tackling the Challenges in Scene Graph Generation with Local-to-Global Interactions","code":"https://github.com/sangminwoo/Local-to-Global-Interaction-Networks-SGG","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"VCTree","metrics":{"Recall@50":"27.9"},"uses_additional_data":false,"paper_date":"2018-12-05","paper":"/paper/learning-to-compose-dynamic-tree-structures","paper_url":"http://arxiv.org/abs/1812.01880v1","paper_title":"Learning to Compose Dynamic Tree Structures for Visual Contexts","code":"https://github.com/KaihuaTang/Scene-Graph-Benchmark.pytorch","n_code_links":6,"syntology":{"n_ran":4,"n_unverified":10,"n_samples":14,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"NODIS","metrics":{"Recall@20":"21.6","Recall@50":"27.7"},"uses_additional_data":false,"paper_date":"2020-01-14","paper":"/paper/nodis-neural-ordinary-differential-scene","paper_url":"https://arxiv.org/abs/2001.04735v3","paper_title":"NODIS: Neural Ordinary Differential Scene Understanding","code":"https://github.com/yrcong/NODIS","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"KERN","metrics":{"Recall@50":"27.1"},"uses_additional_data":false,"paper_date":"2019-03-08","paper":"/paper/knowledge-embedded-routing-network-for-scene","paper_url":"http://arxiv.org/abs/1903.03326v1","paper_title":"Knowledge-Embedded Routing Network for Scene Graph Generation","code":"https://github.com/yuweihao/KERN","n_code_links":3,"syntology":{"n_ran":3,"n_unverified":5,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"DLFE","metrics":{"Recall@50":"25.4"},"uses_additional_data":false,"paper_date":"2021-07-05","paper":"/paper/recovering-the-unbiased-scene-graphs-from-the","paper_url":"https://arxiv.org/abs/2107.02112v1","paper_title":"Recovering the Unbiased Scene Graphs from the Biased Ones","code":"https://github.com/coldmanck/recovering-unbiased-scene-graphs","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"IETrans","metrics":{"Recall@100":"27.2","Recall@50":"23.5","mean Recall @100":"18.0"},"uses_additional_data":false,"paper_date":"2022-03-22","paper":"/paper/fine-grained-scene-graph-generation-with-data","paper_url":"https://arxiv.org/abs/2203.11654v2","paper_title":"Fine-Grained Scene Graph Generation with Data Transfer","code":"https://github.com/waxnkw/ietrans-sgg.pytorch","n_code_links":2,"syntology":null},{"rank_in_archive_order":12,"model":"ADTrans","metrics":{"Recall@50":"23.0","mR@100":"19.2","mR@50":"15.8","mean Recall @100":"19.2","mean Recall @20":"12.3"},"uses_additional_data":false,"paper_date":"2023-07-28","paper":"/paper/panoptic-scene-graph-generation-with","paper_url":"https://arxiv.org/abs/2307.15567v3","paper_title":"Panoptic Scene Graph Generation with Semantics-Prototype Learning","code":"https://github.com/lili0415/psg-biased-annotation","n_code_links":1,"syntology":null},{"rank_in_archive_order":13,"model":"Graph-RCNN","metrics":{"Recall@50":"11.4"},"uses_additional_data":false,"paper_date":"2018-08-01","paper":"/paper/graph-r-cnn-for-scene-graph-generation","paper_url":"http://arxiv.org/abs/1808.00191v1","paper_title":"Graph R-CNN for Scene Graph Generation","code":"https://github.com/jwyang/graph-rcnn.pytorch","n_code_links":3,"syntology":null},{"rank_in_archive_order":14,"model":"MSDN","metrics":{"Recall@50":"10.72"},"uses_additional_data":false,"paper_date":"2017-07-31","paper":"/paper/scene-graph-generation-from-objects-phrases","paper_url":"http://arxiv.org/abs/1707.09700v2","paper_title":"Scene Graph Generation from Objects, Phrases and Region Captions","code":"https://github.com/yikang-li/MSDN","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"C-bias","metrics":{"mean Recall @100":"17.24","mean Recall @20":"11.63"},"uses_additional_data":false,"paper_date":"2022-03-17","paper":"/paper/biasing-like-human-a-cognitive-bias-framework","paper_url":"https://arxiv.org/abs/2203.09160v1","paper_title":"Biasing Like Human: A Cognitive Bias Framework for Scene Graph Generation","code":"https://github.com/rafa-cxg/PySGG-cxg","n_code_links":1,"syntology":null},{"rank_in_archive_order":16,"model":"CogTree","metrics":{"mean Recall @20":"7.9"},"uses_additional_data":false,"paper_date":"2020-09-16","paper":"/paper/cogtree-cognition-tree-loss-for-unbiased","paper_url":"https://arxiv.org/abs/2009.07526v2","paper_title":"CogTree: Cognition Tree Loss for Unbiased Scene Graph Generation","code":"https://github.com/CYVincent/Scene-Graph-Transformer-CogTree","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"ExpressiveSGG","metrics":{"R@100":"39.12"},"uses_additional_data":false,"paper_date":"2022-05-31","paper":"/paper/expressive-scene-graph-generation-using","paper_url":"https://link.springer.com/chapter/10.1007/978-3-031-06981-9_6","paper_title":"Expressive Scene Graph Generation Using Commonsense Knowledge Infusion for Visual Understanding and Reasoning","code":"https://github.com/jaleedkhan/neusire","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"NeuSyRE","metrics":{"R@100":"39.1","mR@100":"12.6","mR@50":"10.9"},"uses_additional_data":false,"paper_date":"2023-11-05","paper":"/paper/neusyre-neuro-symbolic-visual-understanding","paper_url":"https://www.semantic-web-journal.net/content/neusyre-neuro-symbolic-visual-understanding-and-reasoning-framework-based-scene-graph-0","paper_title":"NeuSyRE: Neuro-Symbolic Visual Understanding and Reasoning Framework based on Scene Graph Enrichment","code":"https://github.com/jaleedkhan/neusire","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"KnowZRel","metrics":{"zR@100":"35.65","zR@20":"14.22","zR@50":"25.43"},"uses_additional_data":false,"paper_date":"2025-02-21","paper":"/paper/knowzrel-common-sense-knowledge-based-zero","paper_url":"https://ieeexplore.ieee.org/document/10897903","paper_title":"KnowZRel: Common Sense Knowledge-based Zero-Shot Relationship Retrieval for Generalised Scene Graph Generation","code":"https://github.com/jaleedkhan/zsrr-sgg","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 6,264 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":6,"rows_with_any_sample_ran":6,"distinct_papers_with_graph_line":5,"distinct_papers_with_any_sample_ran":5,"samples_over_distinct_papers":{"n_ran":13,"n_unverified":26,"n_samples":39,"n_pointer_only_licence":9,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":14,"n_unverified":28,"n_samples":42,"n_pointer_only_licence":12,"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"}}}