“…In the last decade variety of methods have been proposed to identify and locate abnormal or prohibited events. The proposed methods can be broadly classified into three major categories i.e., (i) Classical methods [17], [22], [25], [26], [27], [28], (ii) Deep learning methods [4], [6], [7], [8], [29], [30], and (iii) Hybrid methods [13], [21], [31]. A significant research effort has been dedicated towards the manual feature extraction followed by well-established classifiers like one-class support vector machine (OC-SVM) [28], [9], clustering mechanisms like K-nearest neighbor (KNN) [21] and a hardware-friendly classifier KUGDA proposed in [32].…”