Table 2 Statistics about the localized UCF-crime dataset. Number in brackets represent the number of videos in the training set
From: Deep anomaly detection through visual attention in surveillance videos
Type | Value |
|---|---|
Number of videos | Anomalous: 950 (810), Normal: 950 (800) |
Average total number of frames | 13 million |
Dataset length | 128 h |
Anomalous classes tested | Robbery, fighting, road accident |