Summarisation of anomalies in video data

A non-transitory computer readable storage medium storing instructions that, when executed, configure a processor to perform a method of summarising video anomalies in video data, the method comprising:

• receiving video data from one or more video surveillance cameras;
• performing Video Anomaly Detection, VAD, for detecting one or more anomalies in the video data, wherein the VAD computes, for detected objects and/or events in the video data, a respective anomaly score or probability of being anomalous;
• and
• summarising the one or more anomalies by generating at least one composite representation that comprises a background on which one or more visual elements are overlaid, wherein the background represents at least a part of a physical area, and the one or more visual elements respectively provide a preview of at least one anomaly that was detected at a location within the said physical area, the preview comprising visual content extracted from the video data corresponding to the time and location at which the at least one anomaly was detected.

A video processing apparatus, comprising at least one processor configured to:

• receive video data from one or more video surveillance cameras;
• perform Video Anomaly Detection, VAD, for detecting one or more anomalies in the video data, wherein the VAD computes, for detected objects and/or events in the video data, a respective anomaly score or probability of being anomalous;
• and
• summarise the one or more anomalies by generating at least one composite representation that comprises a background on which one or more visual elements are overlaid, wherein the background represents at least a part of a physical area, and the one or more visual elements respectively provide a preview of at least one anomaly that was detected at a location within the said physical area, the preview comprising visual content extracted from the video data corresponding to the time and location at which the at least one anomaly was detected.