{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/self-supervised-transformers-for-unsupervised","title":"Self-Supervised Transformers for Unsupervised Object Discovery using Normalized Cut","arxiv_id":"2202.11539","date":"2022-02-23","proceeding":"CVPR 2022 1","authors":["Yangtao Wang","Xi Shen","Shell Hu","Yuan Yuan","James Crowley","Dominique Vaufreydaz"],"abstract":"Transformers trained with self-supervised learning using self-distillation loss (DINO) have been shown to produce attention maps that highlight salient foreground objects. In this paper, we demonstrate a graph-based approach that uses the self-supervised transformer features to discover an object from an image. Visual tokens are viewed as nodes in a weighted graph with edges representing a connectivity score based on the similarity of tokens. Foreground objects can then be segmented using a normalized graph-cut to group self-similar regions. We solve the graph-cut problem using spectral clustering with generalized eigen-decomposition and show that the second smallest eigenvector provides a cutting solution since its absolute value indicates the likelihood that a token belongs to a foreground object. Despite its simplicity, this approach significantly boosts the performance of unsupervised object discovery: we improve over the recent state of the art LOST by a margin of 6.9%, 8.1%, and 8.1% respectively on the VOC07, VOC12, and COCO20K. The performance can be further improved by adding a second stage class-agnostic detector (CAD). Our proposed method can be easily extended to unsupervised saliency detection and weakly supervised object detection. For unsupervised saliency detection, we improve IoU for 4.9%, 5.2%, 12.9% on ECSSD, DUTS, DUT-OMRON respectively compared to previous state of the art. For weakly supervised object detection, we achieve competitive performance on CUB and ImageNet.","url_abs":"https://arxiv.org/abs/2202.11539v2","url_pdf":"https://arxiv.org/pdf/2202.11539v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"self-supervised-transformers-for-unsupervised","repo_url":"https://github.com/YangtaoWANG95/TokenCut","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-discovery","task_name":"Object Discovery"},{"task_slug":"saliency-detection","task_name":"Saliency Detection"},{"task_slug":"single-object-discovery","task_name":"Single-object discovery"},{"task_slug":"unsupervised-saliency-detection","task_name":"Unsupervised Saliency Detection"},{"task_slug":"weakly-supervised-object-detection","task_name":"Weakly Supervised Object Detection"},{"task_slug":"weakly-supervised-object-localization","task_name":"Weakly-Supervised Object Localization"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"spectral-clustering","method_name":"Spectral Clustering"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/single-object-discovery-on-coco-20k","task":"Single-object discovery","dataset":"COCO_20k","model":"TokenCut + CAD","rank_in_archive_order":4,"of":10,"metrics":{"CorLoc":"62.6"},"uses_additional_data":false},{"leaderboard":"/sota/single-object-discovery-on-coco-20k","task":"Single-object discovery","dataset":"COCO_20k","model":"TokenCut","rank_in_archive_order":5,"of":10,"metrics":{"CorLoc":"58.8"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-saliency-detection-on-dut-omron","task":"Unsupervised Saliency Detection","dataset":"DUT-OMRON","model":"TokenCut","rank_in_archive_order":3,"of":3,"metrics":{"Accuracy":"89.7","IoU":"61.8","maximal F-measure":"69.7"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-saliency-detection-on-duts","task":"Unsupervised Saliency Detection","dataset":"DUTS","model":"TokenCut","rank_in_archive_order":3,"of":3,"metrics":{"Accuracy":"91.4","IoU":"62.4","maximal F-measure":"75.5"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-saliency-detection-on-ecssd","task":"Unsupervised Saliency Detection","dataset":"ECSSD","model":"TokenCut","rank_in_archive_order":3,"of":3,"metrics":{"Accuracy":"93.4","IoU":"77.2","maximal F-measure":"87.4"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-object-localization-on-cub-1","task":"Weakly-Supervised Object Localization","dataset":"CUB","model":"TokenCut","rank_in_archive_order":1,"of":1,"metrics":{"Top-1 Localization Accuracy":"72.9"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-object-localization-on-cub","task":"Weakly-Supervised Object Localization","dataset":"CUB-200-2011","model":"TokenCut","rank_in_archive_order":7,"of":10,"metrics":{"Top-1 Localization Accuracy":"72.9"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-object-localization-on-2","task":"Weakly-Supervised Object Localization","dataset":"ImageNet","model":"TokenCut","rank_in_archive_order":4,"of":6,"metrics":{"GT-known localization accuracy":"65.4","Top-1 Localization Accuracy":"52.3"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2202.11539","atlas_url":"https://app.syntology.ai/?focus=2202.11539","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2202.11539"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/YangtaoWANG95/TokenCut","reach":null}],"summary":{"ran_fixture":1},"by_repo_kind":{"listed":{"samples":1,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"9eb48bfcce51ff49","entry":"mask_color_compose","repo":"YangtaoWANG95/TokenCut","repo_kind":"listed","path":"unsupervised_saliency_detection/get_saliency.py","file_url":"https://github.com/YangtaoWANG95/TokenCut/blob/HEAD/unsupervised_saliency_detection/get_saliency.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"9eb48bfcce51ff49"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}