{"url":"/sota/feature-upsampling-on-imagenet","task":{"name":"Feature Upsampling","url":"/task/feature-upsampling","note":null},"dataset":{"name":"ImageNet","url":"/dataset/imagenet"},"category":"Computer Vision","categories":["Computer Vision","Methodology","Natural Language Processing"],"category_note":null,"description":"Deep features are a cornerstone of computer vision research, capturing image semantics and enabling the community to solve downstream tasks even in the zero- or few-shot regime. However, these features often lack the spatial resolution to directly perform dense prediction tasks like segmentation and depth prediction because models aggressively pool information over large areas. Feature Upsampling aims to recover this missing spatial resolution without impacting the space of the original deep features.","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":["ADCC","Average Drop","Average Increase"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"ADCC":null,"Average Drop":null,"Average Increase":null}},"counts":{"rows":8,"rows_with_code":7,"rows_with_paper_page":7,"rows_dated":7,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"JAFAR","metrics":{"ADCC":"73.3","Average Drop":"17.4","Average Increase":"30.9"},"uses_additional_data":false,"paper_date":"2025-06-10","paper":"/paper/jafar-jack-up-any-feature-at-any-resolution-1","paper_url":"https://arxiv.org/abs/2506.11136v1","paper_title":"JAFAR: Jack up Any Feature at Any Resolution","code":"https://github.com/PaulCouairon/JAFAR","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"FeatUp (JBU)","metrics":{"ADCC":"64.3","Average Drop":"15.3","Average Increase":"24.0"},"uses_additional_data":false,"paper_date":"2024-03-15","paper":"/paper/featup-a-model-agnostic-framework-for","paper_url":"https://arxiv.org/abs/2403.10516v2","paper_title":"FeatUp: A Model-Agnostic Framework for Features at Any Resolution","code":"https://github.com/mhamilton723/FeatUp","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":1,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"Bilinear","metrics":{"ADCC":"61.7","Average Drop":"19.0","Average Increase":"18.5"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"Dysample","metrics":{"ADCC":"61.6","Average Drop":"17.8","Average Increase":"20.0"},"uses_additional_data":false,"paper_date":"2023-08-29","paper":"/paper/learning-to-upsample-by-learning-to-sample","paper_url":"https://arxiv.org/abs/2308.15085v1","paper_title":"Learning to Upsample by Learning to Sample","code":"https://github.com/tiny-smart/dysample","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"CARAFE","metrics":{"ADCC":"59.7","Average Drop":"49.9","Average Increase":"4.8"},"uses_additional_data":false,"paper_date":"2019-05-06","paper":"/paper/carafe-content-aware-reassembly-of-features","paper_url":"https://arxiv.org/abs/1905.02188v3","paper_title":"CARAFE: Content-Aware ReAssembly of FEatures","code":"https://github.com/open-mmlab/mmdetection","n_code_links":3,"syntology":null},{"rank_in_archive_order":6,"model":"ReSFU","metrics":{"ADCC":"59.4","Average Drop":"14.5","Average Increase":"24.0"},"uses_additional_data":false,"paper_date":"2024-07-02","paper":"/paper/a-refreshed-similarity-based-upsampler-for","paper_url":"https://arxiv.org/abs/2407.02283v1","paper_title":"A Refreshed Similarity-based Upsampler for Direct High-Ratio Feature Upsampling","code":"https://github.com/zmhhmz/resfu","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":2,"n_samples":8,"n_pointer_only_licence":8}},{"rank_in_archive_order":7,"model":"SAPA","metrics":{"ADCC":"55.4","Average Drop":"8.5","Average Increase":"32.7"},"uses_additional_data":false,"paper_date":"2022-09-26","paper":"/paper/sapa-similarity-aware-point-affiliation-for","paper_url":"https://arxiv.org/abs/2209.12866v2","paper_title":"SAPA: Similarity-Aware Point Affiliation for Feature Upsampling","code":"https://github.com/poppinace/sapa","n_code_links":2,"syntology":null},{"rank_in_archive_order":8,"model":"LiFT","metrics":{"ADCC":"53.0","Average Drop":"66.9","Average Increase":"8.7"},"uses_additional_data":false,"paper_date":"2024-03-21","paper":"/paper/lift-a-surprisingly-simple-lightweight","paper_url":"https://arxiv.org/abs/2403.14625v2","paper_title":"LiFT: A Surprisingly Simple Lightweight Feature Transform for Dense ViT Descriptors","code":"https://github.com/havrylovv/isegprobe","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}}],"since_archive":{"present":false,"note":"No Syntology-extracted rows are published in this build."},"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":4,"rows_with_any_sample_ran":4,"distinct_papers_with_graph_line":4,"distinct_papers_with_any_sample_ran":4,"samples_over_distinct_papers":{"n_ran":14,"n_unverified":3,"n_samples":17,"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":14,"n_unverified":3,"n_samples":17,"n_pointer_only_licence":8,"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"}}}