Presentation

These pages describe an Anomaly Detection dataset (ADds), a collection of video sequences compiled to provide a  test-set  for comparing anomaly detection recognition algorithms. Manual annotation of ground truth events is also provided.

The dataset is focused on one type of anomalous behavior, namely abandoned (stationary) objects in public scenarios.  The purpose of this dataset is to evaluate the detection performance of an anomaly detection algorithm.

 For more details please refer to the "Content" section.

Related publications (using dataset) :

[1] L. Caro-Campos, J.C. SanMiguel, J.M. Martínez,  “Object-size invariant anomaly detection in video surveillance”

Work partially supported by the Spanish Goverment under project TEC2011-25995 (EventVideo).