{"url":"/dataset/imagenet-c-ood-class-out-of-distribution","name":"ImageNet C-OOD (class-out-of-distribution)","full_name":null,"description_markdown":"This dataset was presented as part of the ICLR 2023 paper 𝘈 𝘧𝘳𝘢𝘮𝘦𝘸𝘰𝘳𝘬 𝘧𝘰𝘳 𝘣𝘦𝘯𝘤𝘩𝘮𝘢𝘳𝘬𝘪𝘯𝘨 𝘊𝘭𝘢𝘴𝘴-𝘰𝘶𝘵-𝘰𝘧-𝘥𝘪𝘴𝘵𝘳𝘪𝘣𝘶𝘵𝘪𝘰𝘯 𝘥𝘦𝘵𝘦𝘤𝘵𝘪𝘰𝘯 𝘢𝘯𝘥 𝘪𝘵𝘴 𝘢𝘱𝘱𝘭𝘪𝘤𝘢𝘵𝘪𝘰𝘯 𝘵𝘰 𝘐𝘮𝘢𝘨𝘦𝘕𝘦𝘵.\r\n\r\nIt is a framework that, based on this dataset (a subset of the ImageNet-21k dataset) is able to generate a C-OOD (AKA open-set recognition) benchmark that covers a variety of difficulty levels. these benchmarks are tailored to the evaluated model. This approach provides a more accurate representation of the model’s own performance.\r\n\r\nThe resulting difficulty levels of our framework allow benchmarking with respect to the difficulty levels most relevant to the task. For example, for a task with a high tolerance for\r\nrisk (e.g., a task for an entertainment application), the performance of a model on a median difficulty level might be more important than on the hardest difficulty level (severity 10). \r\nThe opposite might be true for some applications with a low tolerance for risk (e.g., medical applications), for which one requires the best performance to be attained even if the OOD is very hard to detect (severity 10).\r\nThe paper in which the framework was introduced showed that detection algorithms do not always improve performance on all inputs equally, and could even hurt performance for specific difficulty levels and models. Choosing the combination of (model, detection algorithm) based only on the detection performance on all data may yield sub-optimal results for our specific desired level of difficulty.","description_withheld":null,"homepage":"https://github.com/mdabbah/COOD_benchmarking","introduced_date":"2022-12-01","introduced_date_note":null,"introduced_by":null,"license":null,"modalities":[],"tasks":[{"name":"Classification","url":"/task/classification-1","datasets_with_task":"/datasets/task/classification-1"},{"name":"Out-of-Distribution Detection","url":"/task/out-of-distribution-detection","datasets_with_task":"/datasets/task/out-of-distribution-detection"}],"languages":[],"variants":["ImageNet C-OOD (class-out-of-distribution)"],"data_loaders":[],"num_papers_in_archive":1,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[{"leaderboard":"/sota/classification-on-imagenet-c-ood-class-out-of","task":"Classification","dataset_variant":"ImageNet C-OOD (class-out-of-distribution)","rows":5,"metrics":["Detection AUROC (severity 0)","Detection AUROC (severity 5)","Detection AUROC (severity 10)"],"first_row_in_archive_order":{"model":"ViT-L/32-384 with Max-logit","paper":"/paper/a-framework-for-benchmarking-class-out-of-1","metrics":{"Detection AUROC (severity 0)":"0.9958","Detection AUROC (severity 10)":"0.7748","Detection AUROC (severity 5)":"0.9632"},"code_links":[{"title":"mdabbah/COOD_benchmarking","url":"https://github.com/mdabbah/COOD_benchmarking"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"}],"papers_with_a_benchmark_row":[{"paper":"/paper/a-framework-for-benchmarking-class-out-of-1","title":"A framework for benchmarking class-out-of-distribution detection and its application to ImageNet","date":"2023-02-23","rows_on_this_dataset":5,"code_links":1,"syntology":{"read_at":"2026-09-25T09:33:49+00:00","samples_harvested":15,"samples_ran":13,"samples_unverified":2,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}}],"syntology_totals":{"read_at":"2026-09-25T09:33:49+00:00","papers_with_samples":1,"samples_harvested":15,"samples_ran":13,"samples_unverified":2,"pointer_only_for_licence":0,"papers_with_no_sample_that_ran":0,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}