{"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/exposing-outlier-exposure-what-can-be-learned","title":"Exposing Outlier Exposure: What Can Be Learned From Few, One, and Zero Outlier Images","arxiv_id":"2205.11474","date":"2022-05-23","proceeding":null,"authors":["Philipp Liznerski","Lukas Ruff","Robert A. Vandermeulen","Billy Joe Franks","Klaus-Robert Müller","Marius Kloft"],"abstract":"Due to the intractability of characterizing everything that looks unlike the normal data, anomaly detection (AD) is traditionally treated as an unsupervised problem utilizing only normal samples. However, it has recently been found that unsupervised image AD can be drastically improved through the utilization of huge corpora of random images to represent anomalousness; a technique which is known as Outlier Exposure. In this paper we show that specialized AD learning methods seem unnecessary for state-of-the-art performance, and furthermore one can achieve strong performance with just a small collection of Outlier Exposure data, contradicting common assumptions in the field of AD. We find that standard classifiers and semi-supervised one-class methods trained to discern between normal samples and relatively few random natural images are able to outperform the current state of the art on an established AD benchmark with ImageNet. Further experiments reveal that even one well-chosen outlier sample is sufficient to achieve decent performance on this benchmark (79.3% AUC). We investigate this phenomenon and find that one-class methods are more robust to the choice of training outliers, indicating that there are scenarios where these are still more useful than standard classifiers. Additionally, we include experiments that delineate the scenarios where our results hold. Lastly, no training samples are necessary when one uses the representations learned by CLIP, a recent foundation model, which achieves state-of-the-art AD results on CIFAR-10 and ImageNet in a zero-shot setting.","url_abs":"https://arxiv.org/abs/2205.11474v2","url_pdf":"https://arxiv.org/pdf/2205.11474v2.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":"exposing-outlier-exposure-what-can-be-learned","repo_url":"https://github.com/liznerski/eoe","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"anomaly-detection","task_name":"Anomaly Detection"}],"methods":[{"method_slug":"clip","method_name":"CLIP"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out","task":"Anomaly Detection","dataset":"Leave-One-Class-Out CIFAR-10","model":"BCE-CLIP","rank_in_archive_order":1,"of":6,"metrics":{"AUROC":"98.4"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out","task":"Anomaly Detection","dataset":"Leave-One-Class-Out CIFAR-10","model":"CLIP (zero shot)","rank_in_archive_order":2,"of":6,"metrics":{"AUROC":"92.2"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out","task":"Anomaly Detection","dataset":"Leave-One-Class-Out CIFAR-10","model":"Binary Cross Entropy (OE)","rank_in_archive_order":3,"of":6,"metrics":{"AUROC":"86.6"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out","task":"Anomaly Detection","dataset":"Leave-One-Class-Out CIFAR-10","model":"HSC","rank_in_archive_order":4,"of":6,"metrics":{"AUROC":"84.8"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out","task":"Anomaly Detection","dataset":"Leave-One-Class-Out CIFAR-10","model":"DSAD","rank_in_archive_order":5,"of":6,"metrics":{"AUROC":"84.2"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out","task":"Anomaly Detection","dataset":"Leave-One-Class-Out CIFAR-10","model":"DSVDD","rank_in_archive_order":6,"of":6,"metrics":{"AUROC":"52.2"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out-1","task":"Anomaly Detection","dataset":"Leave-One-Class-Out ImageNet-30","model":"BCE-CLIP (OE)","rank_in_archive_order":1,"of":6,"metrics":{"AUROC":"99.3"},"uses_additional_data":true},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out-1","task":"Anomaly Detection","dataset":"Leave-One-Class-Out ImageNet-30","model":"CLIP (zero shot)","rank_in_archive_order":2,"of":6,"metrics":{"AUROC":"97.8"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out-1","task":"Anomaly Detection","dataset":"Leave-One-Class-Out ImageNet-30","model":"DSAD","rank_in_archive_order":3,"of":6,"metrics":{"AUROC":"88.8"},"uses_additional_data":true},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out-1","task":"Anomaly Detection","dataset":"Leave-One-Class-Out ImageNet-30","model":"HSC (OE)","rank_in_archive_order":4,"of":6,"metrics":{"AUROC":"88.3"},"uses_additional_data":true},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out-1","task":"Anomaly Detection","dataset":"Leave-One-Class-Out ImageNet-30","model":"Binary Cross Entropy (OE)","rank_in_archive_order":5,"of":6,"metrics":{"AUROC":"88.2"},"uses_additional_data":true},{"leaderboard":"/sota/anomaly-detection-on-leave-one-class-out-1","task":"Anomaly Detection","dataset":"Leave-One-Class-Out ImageNet-30","model":"DSVDD","rank_in_archive_order":6,"of":6,"metrics":{"AUROC":"49.7"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-one-class-cifar-10","task":"Anomaly Detection","dataset":"One-class CIFAR-10","model":"CLIP (OE)","rank_in_archive_order":1,"of":36,"metrics":{"AUROC":"99.6"},"uses_additional_data":true},{"leaderboard":"/sota/anomaly-detection-on-one-class-cifar-10","task":"Anomaly Detection","dataset":"One-class CIFAR-10","model":"CLIP (zero shot)","rank_in_archive_order":7,"of":36,"metrics":{"AUROC":"98.5"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-one-class-imagenet-30","task":"Anomaly Detection","dataset":"One-class ImageNet-30","model":"BCE-Clip (OE)","rank_in_archive_order":1,"of":11,"metrics":{"AUROC":"99.90"},"uses_additional_data":true},{"leaderboard":"/sota/anomaly-detection-on-one-class-imagenet-30","task":"Anomaly Detection","dataset":"One-class ImageNet-30","model":"CLIP (Zero Shot)","rank_in_archive_order":2,"of":11,"metrics":{"AUROC":"99.88"},"uses_additional_data":false},{"leaderboard":"/sota/anomaly-detection-on-one-class-imagenet-30","task":"Anomaly Detection","dataset":"One-class ImageNet-30","model":"Binary Cross Entropy (OE)","rank_in_archive_order":3,"of":11,"metrics":{"AUROC":"97.7"},"uses_additional_data":true}],"syntology":{"syntology_url":"https://syntology.ai/paper/2205.11474","atlas_url":"https://app.syntology.ai/?focus=2205.11474","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2205.11474"}},"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/liznerski/eoe","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":1},"by_repo_kind":{"official":{"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":"95ff8c10f19afc85","entry":"global_contrast_normalization","repo":"liznerski/eoe","repo_kind":"official","path":"src/eoe/datasets/bases.py","file_url":"https://github.com/liznerski/eoe/blob/HEAD/src/eoe/datasets/bases.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"95ff8c10f19afc85"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}