{"url":"/sota/open-vocabulary-semantic-segmentation-on-1","task":{"name":"Open Vocabulary Semantic Segmentation","url":"/task/open-vocabulary-semantic-segmentation","note":null},"dataset":{"name":"PASCAL Context-59","url":null},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"Open-vocabulary semantic segmentation models aim to accurately assign a semantic label to each pixel in an image from a set of arbitrary open-vocabulary texts.","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":["mIoU"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"mIoU":null}},"counts":{"rows":24,"rows_with_code":23,"rows_with_paper_page":24,"rows_dated":24,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"HyperSeg","metrics":{"mIoU":"64.6"},"uses_additional_data":true,"paper_date":"2024-11-26","paper":"/paper/hyperseg-towards-universal-visual","paper_url":"https://arxiv.org/abs/2411.17606v2","paper_title":"HyperSeg: Towards Universal Visual Segmentation with Large Language Model","code":"https://github.com/congvvc/HyperSeg","n_code_links":1,"syntology":{"n_ran":7,"n_unverified":10,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":2,"model":"SILC","metrics":{"mIoU":"63.5"},"uses_additional_data":false,"paper_date":"2023-10-20","paper":"/paper/silc-improving-vision-language-pretraining","paper_url":"https://arxiv.org/abs/2310.13355v2","paper_title":"SILC: Improving Vision Language Pretraining with Self-Distillation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"CAT-Seg","metrics":{"mIoU":"63.3"},"uses_additional_data":false,"paper_date":"2023-03-21","paper":"/paper/cat-seg-cost-aggregation-for-open-vocabulary","paper_url":"https://arxiv.org/abs/2303.11797v2","paper_title":"CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation","code":"https://github.com/KU-CVLAB/CAT-Seg","n_code_links":3,"syntology":{"n_ran":5,"n_unverified":1,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":4,"model":"MaskCLIP++","metrics":{"mIoU":"62.5"},"uses_additional_data":false,"paper_date":"2024-12-16","paper":"/paper/maskclip-a-mask-based-clip-fine-tuning","paper_url":"https://arxiv.org/abs/2412.11464v2","paper_title":"MaskCLIP++: A Mask-Based CLIP Fine-tuning Framework for Open-Vocabulary Image Segmentation","code":"https://github.com/hvision-nku/maskclippp","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"CLIPSelf","metrics":{"mIoU":"62.3"},"uses_additional_data":false,"paper_date":"2023-10-02","paper":"/paper/clipself-vision-transformer-distills-itself","paper_url":"https://arxiv.org/abs/2310.01403v2","paper_title":"CLIPSelf: Vision Transformer Distills Itself for Open-Vocabulary Dense Prediction","code":"https://github.com/wusize/clipself","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":4,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":6,"model":"UMG-CLIP-L/14","metrics":{"mIoU":"61.0"},"uses_additional_data":false,"paper_date":"2024-01-12","paper":"/paper/umg-clip-a-unified-multi-granularity-vision","paper_url":"https://arxiv.org/abs/2401.06397v3","paper_title":"UMG-CLIP: A Unified Multi-Granularity Vision Generalist for Open-World Understanding","code":"https://github.com/lygsbw/umg-clip","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"SED","metrics":{"mIoU":"60.6"},"uses_additional_data":false,"paper_date":"2023-11-27","paper":"/paper/sed-a-simple-encoder-decoder-for-open","paper_url":"https://arxiv.org/abs/2311.15537v2","paper_title":"SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentation","code":"https://github.com/xb534/sed","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":1,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"Mask-Adapter","metrics":{"mIoU":"60.4"},"uses_additional_data":false,"paper_date":"2024-12-05","paper":"/paper/mask-adapter-the-devil-is-in-the-masks-for","paper_url":"https://arxiv.org/abs/2412.04533v1","paper_title":"Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation","code":"https://github.com/hustvl/maskadapter","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"EBSeg-L","metrics":{"mIoU":"60.2"},"uses_additional_data":false,"paper_date":"2024-06-14","paper":"/paper/open-vocabulary-semantic-segmentation-with-4","paper_url":"https://arxiv.org/abs/2406.09829v1","paper_title":"Open-Vocabulary Semantic Segmentation with Image Embedding Balancing","code":"https://github.com/slonetime/ebseg","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":9,"n_samples":12,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"MAFT+","metrics":{"mIoU":"59.4"},"uses_additional_data":false,"paper_date":"2024-08-01","paper":"/paper/collaborative-vision-text-representation","paper_url":"https://arxiv.org/abs/2408.00744v2","paper_title":"Collaborative Vision-Text Representation Optimizing for Open-Vocabulary Segmentation","code":"https://github.com/jiaosiyu1999/MAFT-Plus","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"SCAN","metrics":{"mIoU":"59.3"},"uses_additional_data":false,"paper_date":"2023-12-07","paper":"/paper/open-vocabulary-segmentation-with-semantic","paper_url":"https://arxiv.org/abs/2312.04089v2","paper_title":"Open-Vocabulary Segmentation with Semantic-Assisted Calibration","code":"https://github.com/yongliu20/SCAN","n_code_links":2,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"n_pointer_only_licence":10}},{"rank_in_archive_order":12,"model":"MAFT-ViTL","metrics":{"mIoU":"58.5"},"uses_additional_data":false,"paper_date":"2023-09-30","paper":"/paper/learning-mask-aware-clip-representations-for","paper_url":"https://arxiv.org/abs/2310.00240v1","paper_title":"Learning Mask-aware CLIP Representations for Zero-Shot Segmentation","code":"https://github.com/jiaosiyu1999/maft","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":6,"n_samples":7,"n_pointer_only_licence":2}},{"rank_in_archive_order":13,"model":"FC-CLIP","metrics":{"mIoU":"58.4"},"uses_additional_data":false,"paper_date":"2023-08-04","paper":"/paper/convolutions-die-hard-open-vocabulary-1","paper_url":"https://arxiv.org/abs/2308.02487v2","paper_title":"Convolutions Die Hard: Open-Vocabulary Segmentation with Single Frozen Convolutional CLIP","code":"https://github.com/bytedance/fc-clip","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":1,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"ODISE","metrics":{"mIoU":"57.3"},"uses_additional_data":false,"paper_date":"2023-03-08","paper":"/paper/open-vocabulary-panoptic-segmentation-with-1","paper_url":"https://arxiv.org/abs/2303.04803v4","paper_title":"Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models","code":"https://github.com/nvlabs/odise","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"OVSeg Swin-B","metrics":{"mIoU":"55.7"},"uses_additional_data":false,"paper_date":"2022-10-09","paper":"/paper/open-vocabulary-semantic-segmentation-with","paper_url":"https://arxiv.org/abs/2210.04150v3","paper_title":"Open-Vocabulary Semantic Segmentation with Mask-adapted CLIP","code":"https://github.com/facebookresearch/ov-seg","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":1,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":16,"model":"PACL","metrics":{"mIoU":"50.1"},"uses_additional_data":false,"paper_date":"2022-12-09","paper":"/paper/open-vocabulary-semantic-segmentation-with-2","paper_url":"https://arxiv.org/abs/2212.04994v1","paper_title":"Open Vocabulary Semantic Segmentation with Patch Aligned Contrastive Learning","code":"https://github.com/paulcouairon/diffcut","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"SimSeg","metrics":{"mIoU":"47.7"},"uses_additional_data":false,"paper_date":"2021-12-29","paper":"/paper/2112-14757","paper_url":"https://arxiv.org/abs/2112.14757v2","paper_title":"A Simple Baseline for Open-Vocabulary Semantic Segmentation with Pre-trained Vision-language Model","code":"https://github.com/mendelxu/zsseg.baseline","n_code_links":2,"syntology":null},{"rank_in_archive_order":18,"model":"MaskCLIP","metrics":{"mIoU":"45.9"},"uses_additional_data":false,"paper_date":"2022-08-18","paper":"/paper/open-vocabulary-panoptic-segmentation-with","paper_url":"https://arxiv.org/abs/2208.08984v2","paper_title":"Open-Vocabulary Universal Image Segmentation with MaskCLIP","code":"https://github.com/mlpc-ucsd/maskclip","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":19,"model":"TaAlign(trained with image-text pairs)","metrics":{"mIoU":"37.6"},"uses_additional_data":false,"paper_date":"2023-12-21","paper":"/paper/tagalign-improving-vision-language-alignment","paper_url":"https://arxiv.org/abs/2312.14149v4","paper_title":"TagAlign: Improving Vision-Language Alignment with Multi-Tag Classification","code":"https://github.com/Qinying-Liu/TagAlign","n_code_links":1,"syntology":null},{"rank_in_archive_order":20,"model":"TTD (TCL)","metrics":{"mIoU":"37.4"},"uses_additional_data":false,"paper_date":"2024-03-30","paper":"/paper/ttd-text-tag-self-distillation-enhancing","paper_url":"https://arxiv.org/abs/2404.00384v2","paper_title":"TTD: Text-Tag Self-Distillation Enhancing Image-Text Alignment in CLIP to Alleviate Single Tag Bias","code":"https://github.com/shjo-april/TTD","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":21,"model":"LaVG","metrics":{"mIoU":"34.7"},"uses_additional_data":false,"paper_date":"2024-08-09","paper":"/paper/in-defense-of-lazy-visual-grounding-for-open","paper_url":"https://arxiv.org/abs/2408.04961v1","paper_title":"In Defense of Lazy Visual Grounding for Open-Vocabulary Semantic Segmentation","code":"https://github.com/dahyun-kang/lazygrounding","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"TCL","metrics":{"mIoU":"33.9"},"uses_additional_data":false,"paper_date":"2022-12-01","paper":"/paper/learning-to-generate-text-grounded-mask-for","paper_url":"https://arxiv.org/abs/2212.00785v2","paper_title":"Learning to Generate Text-grounded Mask for Open-world Semantic Segmentation from Only Image-Text Pairs","code":"https://github.com/kakaobrain/tcl","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":3,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"TTD (MaskCLIP)","metrics":{"mIoU":"31.0"},"uses_additional_data":false,"paper_date":"2024-03-30","paper":"/paper/ttd-text-tag-self-distillation-enhancing","paper_url":"https://arxiv.org/abs/2404.00384v2","paper_title":"TTD: Text-Tag Self-Distillation Enhancing Image-Text Alignment in CLIP to Alleviate Single Tag Bias","code":"https://github.com/shjo-april/TTD","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":24,"model":"CLIP Surgery (original CLIP without any fine-tuning)","metrics":{"mIoU":"29.3"},"uses_additional_data":false,"paper_date":"2023-04-12","paper":"/paper/clip-surgery-for-better-explainability-with","paper_url":"https://arxiv.org/abs/2304.05653v2","paper_title":"A Closer Look at the Explainability of Contrastive Language-Image Pre-training","code":"https://github.com/xmed-lab/clip_surgery","n_code_links":2,"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":["mIoU"],"entries":[{"paper":"/paper/arxiv-2606-16996","paper_title":"ActiveSAM: Image-Conditional Class Pruning for Fast and Accurate Open-Vocabulary Segmentation","arxiv_id":"2606.16996","model":"ActiveSAM Ours · without background · P59","row_label":"ActiveSAM Ours","configuration":"without background · P59","configuration_in_model":true,"configuration_not_shown_because":null,"values":{"mIoU":"60.1"},"date":"2026-06-15","month_from_arxiv_id":null,"where_in_paper":"Table 1, row “ActiveSAM Ours”","table_label":"1","table_label_from":"the caption's own label","table_position_1based":1,"caption_head":"Table 1: Open-vocabulary semantic segmentation on 8 standard benchmarks.…","cells":[{"column":"mIoU","cell_text":"60.1","column_header":"without background / P59","value":60.1,"table_index_0based":0,"row":31,"col":7}],"code":"4 of 7 ran"}]},"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":13,"rows_with_any_sample_ran":13,"distinct_papers_with_graph_line":12,"distinct_papers_with_any_sample_ran":12,"samples_over_distinct_papers":{"n_ran":37,"n_unverified":41,"n_samples":78,"n_pointer_only_licence":22,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":38,"n_unverified":41,"n_samples":79,"n_pointer_only_licence":23,"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"}}}