{"url":"/sota/face-detection-on-wider-face-hard","task":{"name":"Face Detection","url":"/task/face-detection","note":null},"dataset":{"name":"WIDER Face (Hard)","url":null},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"**Face Detection** is a computer vision task that involves automatically identifying and locating human faces within digital images or videos. It is a fundamental technology that underpins many applications such as face recognition, face tracking, and facial analysis.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [insightface](https://github.com/deepinsight/insightface) )</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":["AP"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"AP":"higher"}},"counts":{"rows":40,"rows_with_code":30,"rows_with_paper_page":40,"rows_dated":40,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"Poly-NL(ResNet-50)","metrics":{"AP":"0.9276"},"uses_additional_data":false,"paper_date":"2021-07-06","paper":"/paper/poly-nl-linear-complexity-non-local-layers","paper_url":"https://arxiv.org/abs/2107.02859v1","paper_title":"Poly-NL: Linear Complexity Non-local Layers with Polynomials","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"ASFD-D6","metrics":{"AP":"0.925"},"uses_additional_data":false,"paper_date":"2022-01-26","paper":"/paper/asfd-automatic-and-scalable-face-detector-1","paper_url":"https://arxiv.org/abs/2201.10781v1","paper_title":"ASFD: Automatic and Scalable Face Detector","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"RetinaFace+Lpts+Lpixel","metrics":{"AP":"0.91857"},"uses_additional_data":false,"paper_date":"2019-05-02","paper":"/paper/190500641","paper_url":"https://arxiv.org/abs/1905.00641v2","paper_title":"RetinaFace: Single-stage Dense Face Localisation in the Wild","code":"https://github.com/deepinsight/insightface","n_code_links":76,"syntology":{"n_ran":64,"n_unverified":27,"n_samples":91,"n_pointer_only_licence":0}},{"rank_in_archive_order":4,"model":"AInnoFace","metrics":{"AP":"0.912"},"uses_additional_data":false,"paper_date":"2019-05-05","paper":"/paper/accurate-face-detection-for-high-performance","paper_url":"https://arxiv.org/abs/1905.01585v3","paper_title":"Accurate Face Detection for High Performance","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":5,"model":"DupNet-L+PACT","metrics":{"AP":"0.906"},"uses_additional_data":false,"paper_date":"2019-11-13","paper":"/paper/dupnet-towards-very-tiny-quantized-cnn-with","paper_url":"https://arxiv.org/abs/1911.05341v1","paper_title":"DupNet: Towards Very Tiny Quantized CNN with Improved Accuracy for Face Detection","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":6,"model":"DSFD","metrics":{"AP":"0.9"},"uses_additional_data":false,"paper_date":"2018-10-24","paper":"/paper/dsfd-dual-shot-face-detector","paper_url":"http://arxiv.org/abs/1810.10220v3","paper_title":"DSFD: Dual Shot Face Detector","code":"https://github.com/Tencent/FaceDetection-DSFD","n_code_links":4,"syntology":null},{"rank_in_archive_order":7,"model":"+ DH + HIM","metrics":{"AP":"0.897"},"uses_additional_data":false,"paper_date":"2018-11-28","paper":"/paper/robust-face-detection-via-learning-small","paper_url":"http://arxiv.org/abs/1811.11662v1","paper_title":"Robust Face Detection via Learning Small Faces on Hard Images","code":"https://github.com/bairdzhang/smallhardface","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"FDNet","metrics":{"AP":"0.896"},"uses_additional_data":false,"paper_date":"2018-02-06","paper":"/paper/face-detection-using-improved-faster-rcnn","paper_url":"http://arxiv.org/abs/1802.02142v1","paper_title":"Face Detection Using Improved Faster RCNN","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":9,"model":"SRN","metrics":{"AP":"0.896"},"uses_additional_data":false,"paper_date":"2018-09-07","paper":"/paper/selective-refinement-network-for-high","paper_url":"http://arxiv.org/abs/1809.02693v1","paper_title":"Selective Refinement Network for High Performance Face Detection","code":"https://github.com/ChiCheng123/SRN","n_code_links":3,"syntology":null},{"rank_in_archive_order":10,"model":"PyramidBox","metrics":{"AP":"0.889"},"uses_additional_data":false,"paper_date":"2018-03-21","paper":"/paper/pyramidbox-a-context-assisted-single-shot","paper_url":"http://arxiv.org/abs/1803.07737v2","paper_title":"PyramidBox: A Context-assisted Single Shot Face Detector","code":"https://github.com/PaddlePaddle/models","n_code_links":5,"syntology":null},{"rank_in_archive_order":11,"model":"Face R-FCN","metrics":{"AP":"0.876"},"uses_additional_data":false,"paper_date":"2017-09-14","paper":"/paper/detecting-faces-using-region-based-fully","paper_url":"http://arxiv.org/abs/1709.05256v2","paper_title":"Detecting Faces Using Region-based Fully Convolutional Networks","code":"https://github.com/vikramkarthikeyan/Face-R-FCN","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"MogFace (HCAM)","metrics":{"AP":"0.874"},"uses_additional_data":false,"paper_date":"2021-03-20","paper":"/paper/mogface-rethinking-scale-augmentation-on-the","paper_url":"https://arxiv.org/abs/2103.11139v5","paper_title":"MogFace: Towards a Deeper Appreciation on Face Detection","code":"https://github.com/damo-cv/mogface","n_code_links":2,"syntology":null},{"rank_in_archive_order":13,"model":"CenterFace","metrics":{"AP":"0.873"},"uses_additional_data":false,"paper_date":"2019-11-09","paper":"/paper/centerface-joint-face-detection-and-alignment","paper_url":"https://arxiv.org/abs/1911.03599v1","paper_title":"CenterFace: Joint Face Detection and Alignment Using Face as Point","code":"https://github.com/Star-Clouds/CenterFace","n_code_links":9,"syntology":null},{"rank_in_archive_order":14,"model":"MogFace (Ali-AMS)","metrics":{"AP":"0.873"},"uses_additional_data":false,"paper_date":"2021-03-20","paper":"/paper/mogface-rethinking-scale-augmentation-on-the","paper_url":"https://arxiv.org/abs/2103.11139v5","paper_title":"MogFace: Towards a Deeper Appreciation on Face Detection","code":"https://github.com/damo-cv/mogface","n_code_links":2,"syntology":null},{"rank_in_archive_order":15,"model":"WSMA-Seg","metrics":{"AP":"0.8723"},"uses_additional_data":false,"paper_date":"2019-04-30","paper":"/paper/segmentation-is-all-you-need","paper_url":"https://arxiv.org/abs/1904.13300v3","paper_title":"Segmentation is All You Need","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":16,"model":"DSFD (RFB)","metrics":{"AP":"0.872"},"uses_additional_data":false,"paper_date":"2018-10-24","paper":"/paper/dsfd-dual-shot-face-detector","paper_url":"http://arxiv.org/abs/1810.10220v3","paper_title":"DSFD: Dual Shot Face Detector","code":"https://github.com/Tencent/FaceDetection-DSFD","n_code_links":4,"syntology":null},{"rank_in_archive_order":17,"model":"YOLOv5x6","metrics":{"AP":"0.8655"},"uses_additional_data":false,"paper_date":"2021-05-27","paper":"/paper/yolo5face-why-reinventing-a-face-detector","paper_url":"https://arxiv.org/abs/2105.12931v3","paper_title":"YOLO5Face: Why Reinventing a Face Detector","code":"https://github.com/deepcam-cn/yolov5-face","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":18,"model":"YOLOv5l6","metrics":{"AP":"0.8588"},"uses_additional_data":false,"paper_date":"2021-05-27","paper":"/paper/yolo5face-why-reinventing-a-face-detector","paper_url":"https://arxiv.org/abs/2105.12931v3","paper_title":"YOLO5Face: Why Reinventing a Face Detector","code":"https://github.com/deepcam-cn/yolov5-face","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":19,"model":"SCRFD-34GF","metrics":{"AP":"0.8529"},"uses_additional_data":false,"paper_date":"2021-05-10","paper":"/paper/sample-and-computation-redistribution-for","paper_url":"https://arxiv.org/abs/2105.04714v1","paper_title":"Sample and Computation Redistribution for Efficient Face Detection","code":"https://github.com/deepinsight/insightface","n_code_links":8,"syntology":{"n_ran":7,"n_unverified":0,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"S3FD(F+S+M)","metrics":{"AP":"0.852"},"uses_additional_data":false,"paper_date":"2017-08-17","paper":"/paper/s3fd-single-shot-scale-invariant-face","paper_url":"http://arxiv.org/abs/1708.05237v3","paper_title":"S$^3$FD: Single Shot Scale-invariant Face Detector","code":"https://github.com/sfzhang15/SFD","n_code_links":3,"syntology":null},{"rank_in_archive_order":21,"model":"YOLOv5m","metrics":{"AP":"0.852"},"uses_additional_data":false,"paper_date":"2021-05-27","paper":"/paper/yolo5face-why-reinventing-a-face-detector","paper_url":"https://arxiv.org/abs/2105.12931v3","paper_title":"YOLO5Face: Why Reinventing a Face Detector","code":"https://github.com/deepcam-cn/yolov5-face","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":22,"model":"EXTD","metrics":{"AP":"0.850"},"uses_additional_data":false,"paper_date":"2019-06-15","paper":"/paper/extd-extremely-tiny-face-detector-via","paper_url":"https://arxiv.org/abs/1906.06579v2","paper_title":"EXTD: Extremely Tiny Face Detector via Iterative Filter Reuse","code":"https://github.com/clovaai/EXTD_Pytorch","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":9,"n_samples":13,"n_pointer_only_licence":2}},{"rank_in_archive_order":23,"model":"YOLOv5l","metrics":{"AP":"0.845"},"uses_additional_data":false,"paper_date":"2021-05-27","paper":"/paper/yolo5face-why-reinventing-a-face-detector","paper_url":"https://arxiv.org/abs/2105.12931v3","paper_title":"YOLO5Face: Why Reinventing a Face Detector","code":"https://github.com/deepcam-cn/yolov5-face","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":24,"model":"img2pose","metrics":{"AP":"0.839"},"uses_additional_data":false,"paper_date":"2020-12-14","paper":"/paper/img2pose-face-alignment-and-detection-via","paper_url":"https://arxiv.org/abs/2012.07791v2","paper_title":"img2pose: Face Alignment and Detection via 6DoF, Face Pose Estimation","code":"https://github.com/vitoralbiero/img2pose","n_code_links":2,"syntology":null},{"rank_in_archive_order":25,"model":"SCRFD-10GF","metrics":{"AP":"0.8305"},"uses_additional_data":false,"paper_date":"2021-05-10","paper":"/paper/sample-and-computation-redistribution-for","paper_url":"https://arxiv.org/abs/2105.04714v1","paper_title":"Sample and Computation Redistribution for Efficient Face Detection","code":"https://github.com/deepinsight/insightface","n_code_links":8,"syntology":{"n_ran":7,"n_unverified":0,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":26,"model":"YOLOv5s","metrics":{"AP":"0.828"},"uses_additional_data":false,"paper_date":"2021-05-27","paper":"/paper/yolo5face-why-reinventing-a-face-detector","paper_url":"https://arxiv.org/abs/2105.12931v3","paper_title":"YOLO5Face: Why Reinventing a Face Detector","code":"https://github.com/deepcam-cn/yolov5-face","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":27,"model":"Massively-large receptive fields","metrics":{"AP":"0.823"},"uses_additional_data":false,"paper_date":"2016-12-13","paper":"/paper/finding-tiny-faces","paper_url":"http://arxiv.org/abs/1612.04402v2","paper_title":"Finding Tiny Faces","code":"https://github.com/peiyunh/tiny","n_code_links":20,"syntology":{"n_ran":0,"n_unverified":14,"n_samples":14,"n_pointer_only_licence":0}},{"rank_in_archive_order":28,"model":"MSCNN","metrics":{"AP":"0.809"},"uses_additional_data":false,"paper_date":"2016-07-25","paper":"/paper/a-unified-multi-scale-deep-convolutional","paper_url":"http://arxiv.org/abs/1607.07155v1","paper_title":"A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection","code":"https://github.com/zhaoweicai/mscnn","n_code_links":1,"syntology":null},{"rank_in_archive_order":29,"model":"SCRFD-2.5GF","metrics":{"AP":"0.7787"},"uses_additional_data":false,"paper_date":"2021-05-10","paper":"/paper/sample-and-computation-redistribution-for","paper_url":"https://arxiv.org/abs/2105.04714v1","paper_title":"Sample and Computation Redistribution for Efficient Face Detection","code":"https://github.com/deepinsight/insightface","n_code_links":8,"syntology":{"n_ran":7,"n_unverified":0,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":30,"model":"LFFD","metrics":{"AP":"0.770"},"uses_additional_data":false,"paper_date":"2019-04-24","paper":"/paper/lffd-a-light-and-fast-face-detector-for-edge","paper_url":"https://arxiv.org/abs/1904.10633v3","paper_title":"LFFD: A Light and Fast Face Detector for Edge Devices","code":"https://github.com/osmr/imgclsmob","n_code_links":15,"syntology":{"n_ran":4,"n_unverified":1,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":31,"model":"RNNPool-Face-C","metrics":{"AP":"0.70"},"uses_additional_data":false,"paper_date":"2020-02-27","paper":"/paper/rnnpool-efficient-non-linear-pooling-for-ram","paper_url":"https://arxiv.org/abs/2002.11921v2","paper_title":"RNNPool: Efficient Non-linear Pooling for RAM Constrained Inference","code":"https://github.com/Microsoft/EdgeML","n_code_links":4,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":32,"model":"SCRFD-0.5GF","metrics":{"AP":"0.6851"},"uses_additional_data":false,"paper_date":"2021-05-10","paper":"/paper/sample-and-computation-redistribution-for","paper_url":"https://arxiv.org/abs/2105.04714v1","paper_title":"Sample and Computation Redistribution for Efficient Face Detection","code":"https://github.com/deepinsight/insightface","n_code_links":8,"syntology":{"n_ran":7,"n_unverified":0,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":33,"model":"CMS-RCNN","metrics":{"AP":"0.643"},"uses_additional_data":false,"paper_date":"2016-06-17","paper":"/paper/cms-rcnn-contextual-multi-scale-region-based","paper_url":"http://arxiv.org/abs/1606.05413v1","paper_title":"CMS-RCNN: Contextual Multi-Scale Region-based CNN for Unconstrained Face Detection","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":34,"model":"Multitask Cascade CNN","metrics":{"AP":"0.607"},"uses_additional_data":false,"paper_date":"2016-04-11","paper":"/paper/joint-face-detection-and-alignment-using","paper_url":"http://arxiv.org/abs/1604.02878v1","paper_title":"Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks","code":"https://github.com/serengil/deepface","n_code_links":42,"syntology":{"n_ran":1,"n_unverified":8,"n_samples":9,"n_pointer_only_licence":3}},{"rank_in_archive_order":35,"model":"LDCF+","metrics":{"AP":"0.564"},"uses_additional_data":false,"paper_date":"2017-01-06","paper":"/paper/to-boost-or-not-to-boost-on-the-limits-of","paper_url":"http://arxiv.org/abs/1701.01692v1","paper_title":"To Boost or Not to Boost? On the Limits of Boosted Trees for Object Detection","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":36,"model":"FD-CNN","metrics":{"AP":"0.51"},"uses_additional_data":false,"paper_date":"2017-01-10","paper":"/paper/fast-deep-convolutional-face-detection-in-the","paper_url":"https://www.researchgate.net/publication/308944615_A_Fast_Deep_Convolutional_Neural_Network_for_Face_Detection_in_Big_Visual_Data","paper_title":"Fast Deep Convolutional Face Detection in the Wild Exploiting Hard Sample Mining","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":37,"model":"Multiscale Cascade CNN","metrics":{"AP":"0.400"},"uses_additional_data":true,"paper_date":"2015-11-20","paper":"/paper/wider-face-a-face-detection-benchmark","paper_url":"http://arxiv.org/abs/1511.06523v1","paper_title":"WIDER FACE: A Face Detection Benchmark","code":"https://github.com/kabrau/FaceDetection","n_code_links":1,"syntology":null},{"rank_in_archive_order":38,"model":"Faceness-WIDER","metrics":{"AP":"0.315"},"uses_additional_data":false,"paper_date":"2015-11-20","paper":"/paper/wider-face-a-face-detection-benchmark","paper_url":"http://arxiv.org/abs/1511.06523v1","paper_title":"WIDER FACE: A Face Detection Benchmark","code":"https://github.com/kabrau/FaceDetection","n_code_links":1,"syntology":null},{"rank_in_archive_order":39,"model":"Two-stage CNN","metrics":{"AP":"0.304"},"uses_additional_data":false,"paper_date":"2015-11-20","paper":"/paper/wider-face-a-face-detection-benchmark","paper_url":"http://arxiv.org/abs/1511.06523v1","paper_title":"WIDER FACE: A Face Detection Benchmark","code":"https://github.com/kabrau/FaceDetection","n_code_links":1,"syntology":null},{"rank_in_archive_order":40,"model":"ACF-WIDER","metrics":{"AP":"0.290"},"uses_additional_data":false,"paper_date":"2014-07-15","paper":"/paper/aggregate-channel-features-for-multi-view","paper_url":"https://arxiv.org/abs/1407.4023v2","paper_title":"Aggregate channel features for multi-view face detection","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":15,"rows_with_any_sample_ran":13,"distinct_papers_with_graph_line":8,"distinct_papers_with_any_sample_ran":6,"samples_over_distinct_papers":{"n_ran":82,"n_unverified":60,"n_samples":142,"n_pointer_only_licence":8,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":111,"n_unverified":60,"n_samples":171,"n_pointer_only_licence":16,"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"}}}