Detecting rare events using Kullback–Leibler divergence: A weakly supervised approach

作者：

Highlights：

• We present a weakly supervised approach for rare event detection.

• Coarse annotation, denoting only roughly when an event occurs is needed.

• The approach leverages the rare nature of the target events to its advantage.

• We demonstrate the proposed approach on the popular MIT traffic dataset.

• State-of-the-art performance is shown, alongside being real-time capable.

摘要

•We present a weakly supervised approach for rare event detection.•Coarse annotation, denoting only roughly when an event occurs is needed.•The approach leverages the rare nature of the target events to its advantage.•We demonstrate the proposed approach on the popular MIT traffic dataset.•State-of-the-art performance is shown, alongside being real-time capable.

论文关键词：Event detection,Weakly supervised learning,Kullback–Leibler divergence,Anomaly detection

论文评审过程：Received 14 August 2015, Revised 20 January 2016, Accepted 21 January 2016, Available online 29 January 2016, Version of Record 13 February 2016.

论文官网地址：https://doi.org/10.1016/j.eswa.2016.01.035