{"url":"/sota/crowd-counting-on-shanghaitech-a","task":{"name":"Crowd Counting","url":"/task/crowd-counting","note":null},"dataset":{"name":"ShanghaiTech A","url":"/dataset/shanghaitech"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"**Crowd Counting** is a task to count people in image. It is mainly used in real-life for automated public monitoring such as surveillance and traffic control. Different from object detection, Crowd Counting aims at recognizing arbitrarily sized targets in various situations including sparse and cluttering scenes at the same time.\r\n\r\n\r\n<span class=\"description-source\">Source: [Deep Density-aware Count Regressor ](https://arxiv.org/abs/1908.03314)</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":["MAE","MSE","RMSE"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"MAE":"lower","MSE":"lower","RMSE":"lower"}},"counts":{"rows":35,"rows_with_code":27,"rows_with_paper_page":35,"rows_dated":35,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"EBC-ZIP-B","metrics":{"MAE":"47.81","RMSE":"75.04"},"uses_additional_data":false,"paper_date":"2025-06-24","paper":"/paper/ebc-zip-improving-blockwise-crowd-counting","paper_url":"https://arxiv.org/abs/2506.19955v1","paper_title":"EBC-ZIP: Improving Blockwise Crowd Counting with Zero-Inflated Poisson Regression","code":"https://github.com/yiming-m/ebc-zip","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"APGCC","metrics":{"MAE":"48.8","MSE":"76.7"},"uses_additional_data":false,"paper_date":"2024-05-17","paper":"/paper/improving-point-based-crowd-counting-and","paper_url":"https://arxiv.org/abs/2405.10589v1","paper_title":"Improving Point-based Crowd Counting and Localization Based on Auxiliary Point Guidance","code":"https://github.com/AaronCIH/APGCC","n_code_links":1,"syntology":null},{"rank_in_archive_order":3,"model":"PSL-Net","metrics":{"MAE":"49.9","RMSE":"77.6"},"uses_additional_data":false,"paper_date":"2024-05-13","paper":"/paper/crowd-counting-and-individual-localization","paper_url":"https://ieeexplore.ieee.org/document/10530038","paper_title":"Crowd Counting and Individual Localization Using Pseudo Square Label","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"FGENet","metrics":{"MAE":"51.66","MSE":"85.00"},"uses_additional_data":false,"paper_date":"2024-01-02","paper":"/paper/fgenet-fine-grained-extraction-network-for-2","paper_url":"https://arxiv.org/abs/2401.01208v1","paper_title":"FGENet: Fine-Grained Extraction Network for Congested Crowd Counting","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":5,"model":"VMambaCC","metrics":{"MAE":"51.87","MSE":"81.3"},"uses_additional_data":false,"paper_date":"2024-05-07","paper":"/paper/vmambacc-a-visual-state-space-model-for-crowd","paper_url":"https://arxiv.org/abs/2405.03978v1","paper_title":"VMambaCC: A Visual State Space Model for Crowd Counting","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":6,"model":"CLIP-EBC (ViT-B/16)","metrics":{"MAE":"52.5","RMSE":"85.9"},"uses_additional_data":false,"paper_date":"2024-03-14","paper":"/paper/clip-ebc-clip-can-count-accurately-through","paper_url":"https://arxiv.org/abs/2403.09281v3","paper_title":"CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification","code":"https://github.com/Yiming-M/CLIP-EBC","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"P2PNet","metrics":{"MAE":"52.74","MSE":"85.06"},"uses_additional_data":false,"paper_date":"2021-07-27","paper":"/paper/rethinking-counting-and-localization-in","paper_url":"https://arxiv.org/abs/2107.12746v3","paper_title":"Rethinking Counting and Localization in Crowds:A Purely Point-Based Framework","code":"https://github.com/tencentyouturesearch/crowdcounting-p2pnet","n_code_links":3,"syntology":null},{"rank_in_archive_order":8,"model":"CLIP-EBC (ResNet50)","metrics":{"MAE":"54.0","RMSE":"83.2"},"uses_additional_data":false,"paper_date":"2024-03-14","paper":"/paper/clip-ebc-clip-can-count-accurately-through","paper_url":"https://arxiv.org/abs/2403.09281v3","paper_title":"CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification","code":"https://github.com/Yiming-M/CLIP-EBC","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"GauNet (ResNet-50)","metrics":{"MAE":"54.8","MSE":"89.1"},"uses_additional_data":false,"paper_date":"2022-06-10","paper":"/paper/rethinking-spatial-invariance-of-1","paper_url":"https://arxiv.org/abs/2206.05253v2","paper_title":"Rethinking Spatial Invariance of Convolutional Networks for Object Counting","code":"https://github.com/zhiqic/rethinking-counting","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"LoViTCrowd","metrics":{"MAE":"54.8","RMSE":"80.9"},"uses_additional_data":false,"paper_date":"2022-09-30","paper":"/paper/improving-local-features-with-relevant","paper_url":"https://bmvc2022.mpi-inf.mpg.de/729/","paper_title":"Improving Local Features with Relevant Spatial Information by Vision Transformer for Crowd Counting","code":"https://github.com/nguyen1312/LoViTCrowd","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"M-SFANet+M-SegNet","metrics":{"MAE":"57.55","MSE":"94.48"},"uses_additional_data":false,"paper_date":"2020-03-12","paper":"/paper/encoder-decoder-based-convolutional-neural","paper_url":"https://arxiv.org/abs/2003.05586v5","paper_title":"Encoder-Decoder Based Convolutional Neural Networks with Multi-Scale-Aware Modules for Crowd Counting","code":"https://github.com/Pongpisit-Thanasutives/Variations-of-SFANet-for-Crowd-Counting","n_code_links":2,"syntology":null},{"rank_in_archive_order":12,"model":"SGANet + CL","metrics":{"MAE":"57.6"},"uses_additional_data":false,"paper_date":"2019-11-18","paper":"/paper/segmentation-guided-attention-network-for","paper_url":"https://arxiv.org/abs/1911.07990v2","paper_title":"Crowd Counting via Segmentation Guided Attention Networks and Curriculum Loss","code":"https://github.com/hellowangqian/sganet-crowd-counting","n_code_links":1,"syntology":null},{"rank_in_archive_order":13,"model":"SGANet","metrics":{"MAE":"58"},"uses_additional_data":false,"paper_date":"2019-11-18","paper":"/paper/segmentation-guided-attention-network-for","paper_url":"https://arxiv.org/abs/1911.07990v2","paper_title":"Crowd Counting via Segmentation Guided Attention Networks and Curriculum Loss","code":"https://github.com/hellowangqian/sganet-crowd-counting","n_code_links":1,"syntology":null},{"rank_in_archive_order":14,"model":"S-DCNet","metrics":{"MAE":"58.3"},"uses_additional_data":false,"paper_date":"2019-08-15","paper":"/paper/from-open-set-to-closed-set-counting-objects","paper_url":"https://arxiv.org/abs/1908.06473v1","paper_title":"From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer","code":"https://github.com/xhp-hust-2018-2011/S-DCNet","n_code_links":5,"syntology":{"n_ran":3,"n_unverified":5,"n_samples":8,"n_pointer_only_licence":3}},{"rank_in_archive_order":15,"model":"SPANet","metrics":{"MAE":"59.4","MSE":"92.5"},"uses_additional_data":false,"paper_date":"2019-09-16","paper":"/paper/learning-spatial-awareness-to-improve-crowd","paper_url":"https://arxiv.org/abs/1909.07057v1","paper_title":"Learning Spatial Awareness to Improve Crowd Counting","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":16,"model":"DM-Count","metrics":{"MAE":"59.7"},"uses_additional_data":false,"paper_date":"2020-09-28","paper":"/paper/distribution-matching-for-crowd-counting","paper_url":"https://arxiv.org/abs/2009.13077v2","paper_title":"Distribution Matching for Crowd Counting","code":"https://github.com/cvlab-stonybrook/DM-Count","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":1,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":17,"model":"FusionCount","metrics":{"MAE":"62.2","MSE":"101.2"},"uses_additional_data":false,"paper_date":"2022-02-28","paper":"/paper/fusioncount-efficient-crowd-counting-via","paper_url":"https://arxiv.org/abs/2202.13660v1","paper_title":"FusionCount: Efficient Crowd Counting via Multiscale Feature Fusion","code":"https://github.com/YimingMa/FusionCount","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"DMCount-EBC","metrics":{"MAE":"62.3","RMSE":"98.9"},"uses_additional_data":false,"paper_date":"2024-03-14","paper":"/paper/clip-ebc-clip-can-count-accurately-through","paper_url":"https://arxiv.org/abs/2403.09281v3","paper_title":"CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification","code":"https://github.com/Yiming-M/CLIP-EBC","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"CAN","metrics":{"MAE":"62.3"},"uses_additional_data":false,"paper_date":"2018-11-26","paper":"/paper/context-aware-crowd-counting","paper_url":"http://arxiv.org/abs/1811.10452v2","paper_title":"Context-Aware Crowd Counting","code":"https://github.com/weizheliu/Context-Aware-Crowd-Counting","n_code_links":3,"syntology":{"n_ran":1,"n_unverified":1,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"OrdinalEntropy","metrics":{"MAE":"65.6","MSE":"105.0"},"uses_additional_data":false,"paper_date":"2023-01-21","paper":"/paper/improving-deep-regression-with-ordinal","paper_url":"https://arxiv.org/abs/2301.08915v3","paper_title":"Improving Deep Regression with Ordinal Entropy","code":"https://github.com/needylove/ordinalentropy","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":1,"n_samples":5,"n_pointer_only_licence":5}},{"rank_in_archive_order":21,"model":"CSRNet-EBC","metrics":{"MAE":"66.3","RMSE":"105.0"},"uses_additional_data":false,"paper_date":"2024-03-14","paper":"/paper/clip-ebc-clip-can-count-accurately-through","paper_url":"https://arxiv.org/abs/2403.09281v3","paper_title":"CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification","code":"https://github.com/Yiming-M/CLIP-EBC","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"LSC-CNN","metrics":{"MAE":"66.4"},"uses_additional_data":false,"paper_date":"2019-06-18","paper":"/paper/locate-size-and-count-accurately-resolving","paper_url":"https://arxiv.org/abs/1906.07538v3","paper_title":"Locate, Size and Count: Accurately Resolving People in Dense Crowds via Detection","code":"https://github.com/val-iisc/lsc-cnn","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":9,"n_samples":12,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"SANet","metrics":{"MAE":"67.0"},"uses_additional_data":false,"paper_date":"2018-09-01","paper":"/paper/scale-aggregation-network-for-accurate-and","paper_url":"http://openaccess.thecvf.com/content_ECCV_2018/html/Xinkun_Cao_Scale_Aggregation_Network_ECCV_2018_paper.html","paper_title":"Scale Aggregation Network for Accurate and Efficient Crowd Counting","code":"https://github.com/ZhengPeng7/SANet-Keras","n_code_links":1,"syntology":null},{"rank_in_archive_order":24,"model":"CSRNet","metrics":{"MAE":"68.2"},"uses_additional_data":false,"paper_date":"2018-02-27","paper":"/paper/csrnet-dilated-convolutional-neural-networks","paper_url":"http://arxiv.org/abs/1802.10062v4","paper_title":"CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes","code":"https://github.com/leeyeehoo/CSRNet-pytorch","n_code_links":11,"syntology":{"n_ran":5,"n_unverified":10,"n_samples":15,"n_pointer_only_licence":5}},{"rank_in_archive_order":25,"model":"ic-CNN","metrics":{"MAE":"68.5"},"uses_additional_data":false,"paper_date":"2018-07-26","paper":"/paper/iterative-crowd-counting","paper_url":"http://arxiv.org/abs/1807.09959v1","paper_title":"Iterative Crowd Counting","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":26,"model":"IG-CNN","metrics":{"MAE":"72.5"},"uses_additional_data":false,"paper_date":"2018-07-26","paper":"/paper/divide-and-grow-capturing-huge-diversity-in","paper_url":"http://arxiv.org/abs/1807.09993v1","paper_title":"Divide and Grow: Capturing Huge Diversity in Crowd Images with Incrementally Growing CNN","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":27,"model":"D-ConvNet","metrics":{"MAE":"73.5"},"uses_additional_data":false,"paper_date":"2018-06-01","paper":"/paper/crowd-counting-with-deep-negative-correlation","paper_url":"http://openaccess.thecvf.com/content_cvpr_2018/html/Shi_Crowd_Counting_With_CVPR_2018_paper.html","paper_title":"Crowd Counting With Deep Negative Correlation Learning","code":"https://github.com/shizenglin/Deep-NCL","n_code_links":1,"syntology":null},{"rank_in_archive_order":28,"model":"CP-CNN","metrics":{"MAE":"73.6"},"uses_additional_data":false,"paper_date":"2017-08-02","paper":"/paper/generating-high-quality-crowd-density-maps","paper_url":"http://arxiv.org/abs/1708.00953v1","paper_title":"Generating High-Quality Crowd Density Maps using Contextual Pyramid CNNs","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":29,"model":"Liu et al.","metrics":{"MAE":"73.6"},"uses_additional_data":false,"paper_date":"2018-03-08","paper":"/paper/leveraging-unlabeled-data-for-crowd-counting","paper_url":"http://arxiv.org/abs/1803.03095v1","paper_title":"Leveraging Unlabeled Data for Crowd Counting by Learning to Rank","code":"https://github.com/xialeiliu/CrowdCountingCVPR18","n_code_links":1,"syntology":null},{"rank_in_archive_order":30,"model":"SAFECount","metrics":{"MAE":"73.70"},"uses_additional_data":false,"paper_date":"2022-01-22","paper":"/paper/iterative-correlation-based-feature","paper_url":"https://arxiv.org/abs/2201.08959v5","paper_title":"Few-shot Object Counting with Similarity-Aware Feature Enhancement","code":"https://github.com/zhiyuanyou/SAFECount","n_code_links":1,"syntology":null},{"rank_in_archive_order":31,"model":"ACSCP","metrics":{"MAE":"75.7"},"uses_additional_data":false,"paper_date":"2018-06-01","paper":"/paper/crowd-counting-via-adversarial-cross-scale","paper_url":"http://openaccess.thecvf.com/content_cvpr_2018/html/Shen_Crowd_Counting_via_CVPR_2018_paper.html","paper_title":"Crowd Counting via Adversarial Cross-Scale Consistency Pursuit","code":"https://github.com/RQuispeC/pytorch-ACSCP","n_code_links":1,"syntology":null},{"rank_in_archive_order":32,"model":"Switch-CNN","metrics":{"MAE":"90.4"},"uses_additional_data":false,"paper_date":"2017-08-01","paper":"/paper/switching-convolutional-neural-network-for","paper_url":"http://arxiv.org/abs/1708.00199v2","paper_title":"Switching Convolutional Neural Network for Crowd Counting","code":"https://github.com/val-iisc/crowd-counting-scnn","n_code_links":1,"syntology":null},{"rank_in_archive_order":33,"model":"Cascaded-MTL","metrics":{"MAE":"101.3","MSE":"152.4"},"uses_additional_data":false,"paper_date":"2017-07-30","paper":"/paper/cnn-based-cascaded-multi-task-learning-of","paper_url":"http://arxiv.org/abs/1707.09605v2","paper_title":"CNN-based Cascaded Multi-task Learning of High-level Prior and Density Estimation for Crowd Counting","code":"https://github.com/svishwa/crowdcount-cascaded-mtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":34,"model":"MCNN","metrics":{"MAE":"110.2"},"uses_additional_data":false,"paper_date":"2016-01-01","paper":"/paper/single-image-crowd-counting-via-multi-column-1","paper_url":"https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Zhang_Single-Image_Crowd_Counting_CVPR_2016_paper.pdf","paper_title":"Single-Image Crowd Counting via Multi-Column Convolutional Neural Network","code":"https://github.com/svishwa/crowdcount-mcnn","n_code_links":4,"syntology":null},{"rank_in_archive_order":35,"model":"Zhang et al.","metrics":{"MAE":"181.8"},"uses_additional_data":false,"paper_date":"2015-06-01","paper":"/paper/cross-scene-crowd-counting-via-deep","paper_url":"http://openaccess.thecvf.com/content_cvpr_2015/html/Zhang_Cross-Scene_Crowd_Counting_2015_CVPR_paper.html","paper_title":"Cross-Scene Crowd Counting via Deep Convolutional Neural Networks","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":6,"rows_with_any_sample_ran":6,"distinct_papers_with_graph_line":6,"distinct_papers_with_any_sample_ran":6,"samples_over_distinct_papers":{"n_ran":20,"n_unverified":27,"n_samples":47,"n_pointer_only_licence":13,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":20,"n_unverified":27,"n_samples":47,"n_pointer_only_licence":13,"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"}}}