Evidence catalog
Datasets shape what anomaly systems can claim.
Filter representative records by annotation and task. Every row carries a verification note.
| Dataset | Year / domain | Scale | Temporal | Spatial | Semantic | Tasks |
|---|---|---|---|---|---|---|
| Subway EntranceDetection-Oriented Sourced from survey PDF, Table 3 | 2008Streetscape | 86,535 frames | Frame | None | None | Detection, Localization |
| Subway ExitDetection-Oriented Sourced from survey PDF, Table 3 | 2008Streetscape | 38,940 frames | Frame | None | None | Detection, Localization |
| UMNDetection-Oriented Sourced from survey PDF, Table 3 | 2009Crowd behavior | 3,855 frames | Frame | None | None | Detection, Localization |
| UCSD Ped1Detection-Oriented Sourced from survey PDF, Table 3 | 2013Streetscape | 14,000 frames | Frame | Bounding box | None | Detection, Localization |
| UCSD Ped2Detection-Oriented Sourced from survey PDF, Table 3 | 2013Streetscape | 4,560 frames | Frame | Bounding box | None | Detection, Localization |
| CUHK AvenueDetection-Oriented Sourced from survey PDF, Table 3 | 2013Streetscape | 30,652 frames | Frame | Bounding box | None | Detection, Localization |
| NWPU CampusDetection-Oriented Sourced from survey PDF, Table 3 | 2023Streetscape | 1,466,073 frames | Frame | None | None | Detection, Localization |
| ShanghaiTechDetection-Oriented Sourced from survey PDF, Table 3 | 2017Streetscape | 317,398 frames | Frame | Bounding box | None | Detection, Localization |
| Street SceneDetection-Oriented Sourced from survey PDF, Table 3 | 2020Streetscape | 203,257 frames | Track | Bounding box | None | Detection, Localization |
| ComplexVADDetection-Oriented Sourced from survey PDF, Table 3 | 2025Streetscape | 3,681,438 frames | Track | Bounding box | None | Detection, Localization |
| UCF-CrimeCategory-Oriented Sourced from survey PDF, Table 3 | 2018Crime | 13,741,393 frames | Video | None | Category labels | Detection, Classification |
| UCF-Crime ExtensionCategory-Oriented Sourced from survey PDF, Table 3 | 2021Crime | 14,475,793 frames | Video | None | Category labels | Detection, Classification |
| XD-ViolenceCategory-Oriented Sourced from survey PDF, Table 3 | 2020Violence | 114,096 frames | Video | None | Category labels | Detection, Classification |
| TADCategory-Oriented Sourced from survey PDF, Table 3 | 2024Traffic | 721,280 frames | Frame | Bounding box | Category labels | Detection, Classification |
| BOSSCategory-Oriented Sourced from survey PDF, Table 3 | 2017Multiple | 48,624 frames | Video | None | Category labels | Detection, Classification |
| CamNuvemCategory-Oriented Sourced from survey PDF, Table 3 | 2022Robbery | 6,151,788 frames | Video | None | Category labels | Detection, Classification |
| UbnormalCategory-Oriented Sourced from survey PDF, Table 3 | 2022Multiple | 236,902 frames | Frame | Pixel | Category labels | Detection, Classification |
| SENSE-VADCategory-Oriented Sourced from survey PDF, Table 3 | 2026Autonomous driving | 540,888 frames | Frame | Bounding box | Category labels | Detection, Classification |
| MSADCategory-Oriented Sourced from survey PDF, Table 3 | 2024Multiple | 447,236 frames | Frame | None | Category labels | Detection, Classification |
| CUVAAnomaly-Understanding Sourced from survey PDF, Table 3 | 2024Multiple | 3,345,097 frames | Time duration | None | Video QA | Explanation, Question answering |
| ECVAAnomaly-Understanding Sourced from survey PDF, Table 3 | 2024Multiple | 19,042,560 frames | Time duration | None | Video QA | Explanation, Question answering |
| VANE-BenchAnomaly-Understanding Sourced from survey PDF, Table 3 | 2025Multiple | 951,482 frames | Video | None | Conversational QA | Explanation, Question answering |
| VAGUAnomaly-Understanding Sourced from survey PDF, Table 3 | 2025Multiple | 20,400,000 frames | Period | None | Grounding + QA | Explanation, Question answering, Grounding |
| FineVAUAnomaly-Understanding Sourced from survey PDF, Table 3 | 2026Surveillance | Not reported | None | None | Fine-grained QA | Explanation, Question answering |
| SVTAAnomaly-Understanding Sourced from survey PDF, Table 3 | 2025Multiple | 1,360,000 frames | Video-level | None | Retrieval text pairs | Retrieval, Explanation |
| CVACBenchAnomaly-Understanding Sourced from survey PDF, Table 3 | 2025Surveillance | 60,160 frames | Frame | None | Captioning | Captioning, Explanation |
| HVAU-70KAnomaly-Understanding Sourced from survey PDF, Table 3 | 2025Multiple | 13,855,489 frames | Frame | None | Video QA | Explanation, Question answering |
| UCAAnomaly-Understanding Sourced from survey PDF, Table 3 | 2024Crime | 11,817,597 frames | Frame | None | Video QA | Explanation, Question answering |