OV-CAVED_UCF-Crime_dataset

OV-CAVED UCF-Crime is a benchmark for Open Vocabulary Context Aware Video Event Detection, derived from the UCF-Crime video anomaly detection dataset.

The dataset extends conventional temporal anomaly annotations with the additional information required by the OV-CAVED framework. Each video is associated with an operational context describing the monitored scene and with natural-language event queries that specify the events to be verified. Queries are provided at different levels of granularity and are aligned with temporal supports, enabling query-conditioned evaluation at both chunk level and video level.

The benchmark is designed to support the study of context-aware and open-vocabulary surveillance event verification. Instead of evaluating only whether a video segment is anomalous, OV-CAVED UCF-Crime allows models to be tested on whether a specific user-defined event is visually present under the operational conditions of the scene.

This resource accompanies the paper “Open Vocabulary Context Aware Video Event Detection”, submitted to Information Fusion. The dataset will be uploaded and made publicly available on Zenodo after acceptance of the paper. It is intended for research purposes and supports reproducible evaluation of models that jointly exploit video evidence, operational context, and natural-language event descriptions.