VADNET: INNOVATIVE DEEP LEARNING FRAMEWORK FOR DETECTING AND CLASSIFYING ANOMALIES IN PUBLICSURVEILLANCE FOOTAGE

In addition to aiding in the establishment of evidence about specific instances, in urban areas and other noteworthy public locations, the deployment of public observation cameras has decreased crime.As a result, the field of video analytics research has become crucial.Real-time video analysis and storage are made feasible by technology advancements like cloud computing, AI, and IoT.Simultaneously, the development of deep learning algorithms has made it possible to interpret visual material and perform effective video analytics.The identification of anomalies in surveillance footage has gained importance.Due to the fact that CNNs cannot provide maximum performance unless they are specifically tailored to handle a certain task, the current research in this field is limited.To accomplish this aim, we developed a unique deep learning architecture for the purpose of detecting abnormalities from surveillance films in this work called VADNet.A CNN variation called VADNet is intended to maximize detection performance.In order to effectively detect video abnormalities, we proposed an algorithm known as Intelligent Video Anomaly Detection (IVAD) which makes use of VADNet.The benchmark dataset for our empirical analysis is UCF-Crime.Based on our experimental findings, VADNet achieves the best accuracy of 95.64% when compared to other CNN versions, such as MobileNetV1, ResNet50, and VGG19 models.