Research Article

Integrated Multiscale Appearance Features and Motion Information Prediction Network for Anomaly Detection

Figure 9

Frame-level evaluation results on the recorded real video. (a) Ground truth of the 432th frame (A) and 496th frame (B) labelled as abnormal. (b) Difference between the ground truth and the corresponding predicted frame. (c) Relationship between the recorded real video frames and the regular score.
(a)
(b)
(c)