{"url":"/dataset/celeba-spoof-enroll","name":"CelebA-Spoof-Enroll","full_name":null,"description_markdown":"CelebA-Spoof is a large-scale face anti-spoofing dataset\r\nrecently introduced in [53]. The dataset contains 625,537\r\nimages of 10,177 celebrities captured under different spoof\r\nmediums, environments and illumination conditions. The\r\noriginal dataset proposes three different evaluation protocols. For our experimentation, we focus on the most general\r\n”intra” protocol, in which different spoof types, environments and illumination conditions are used for both training\r\nand testing.\r\n\r\nTo generate CelebA-Spoof-Enroll, the personalized version of the CelebA-Spoof anti-spoofing dataset, we start by\r\nsetting the desired enrollment set size N. We decide for a\r\nconstant number of enrollment images per user to be consistent with the implementation in typical commercial applications and to simplify the dataset definition. For both the\r\ntraining and test split, we count the number of live samples\r\nper user and discard those users having ≤ N live samples.\r\nSo, if we desire a higher value of N, a larger number of\r\nusers would get rejected and hence the number of training\r\nsamples would be less. Note that it is not possible to include\r\nall original data, as a number of users in CelebA-Spoof are\r\nmissing live samples. Nevertheless, only a very small percentage of training and test data is discarded through this\r\nprocess when choosing N < 10.\r\nFor each accepted user, the first N live samples (ordered according to the ascending alphanumeric ordering of\r\nthe original filenames) are chosen to define its enrollment\r\nset. The rest of the live samples and the spoof samples are\r\nmarked as query samples. It is important to note that using\r\nthis deterministic method of obtaining enrollment samples\r\ndoes not introduce unwanted bias as most of the CelebASpoof images are randomly crawled from the internet and\r\nare not ordered according to specific criteria. With this setting, we associate a user identifier with each query sample\r\nsuch that queries can be easily mapped to the correct enrollment set. This method of filtering users for a desired enrollment size N is performed for both training and test split. In\r\nthe rest of the paper, we refer to CelebA-Spoof-EnrollN or\r\nin short CASp-EnrollN to describe the personalized version\r\nof the dataset with N enrollment images\r\n\r\n(See paper for additional details)","description_withheld":null,"homepage":"","introduced_date":"2022-01-05","introduced_date_note":null,"introduced_by":{"paper":"/paper/a-personalized-benchmark-for-face-anti","title":"A personalized benchmark for face anti-spoofing","first_author":"Davide Belli","url":null},"license":null,"modalities":[{"name":"Images","url":"/datasets/modality/images"}],"tasks":[{"name":"Face Anti-Spoofing","url":"/task/face-anti-spoofing","datasets_with_task":"/datasets/task/face-anti-spoofing"}],"languages":[],"variants":["CelebA-Spoof-Enroll","CelebA-Spoof-Enroll5"],"data_loaders":[],"num_papers_in_archive":3,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[{"leaderboard":"/sota/face-anti-spoofing-on-celeba-spoof-enroll5","task":"Face Anti-Spoofing","dataset_variant":"CelebA-Spoof-Enroll5","rows":5,"metrics":["AUC"],"first_row_in_archive_order":{"model":"ResNet 18 Personalized","paper":"/paper/a-personalized-benchmark-for-face-anti","metrics":{"AUC":"99.2"},"code_links":[{"title":"FaceOnLive/Face-Liveness-Detection-SDK-Linux","url":"https://github.com/FaceOnLive/Face-Liveness-Detection-SDK-Linux"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"}],"papers_with_a_benchmark_row":[{"paper":"/paper/a-personalized-benchmark-for-face-anti","title":"A personalized benchmark for face anti-spoofing","date":"2022-01-05","rows_on_this_dataset":3,"code_links":1,"syntology":null},{"paper":"/paper/190409290","title":"FeatherNets: Convolutional Neural Networks as Light as Feather for Face Anti-spoofing","date":"2019-04-22","rows_on_this_dataset":1,"code_links":3,"syntology":null},{"paper":"/paper/very-deep-convolutional-networks-for-large","title":"Very Deep Convolutional Networks for Large-Scale Image Recognition","date":"2014-09-04","rows_on_this_dataset":1,"code_links":305,"syntology":{"read_at":"2026-09-25T09:33:49+00:00","samples_harvested":122,"samples_ran":14,"samples_unverified":108,"pointer_only_for_licence":4,"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":122,"samples_ran":14,"samples_unverified":108,"pointer_only_for_licence":4,"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."}