Suspicious Activity Detection Using Machine Learning

Abstract: Video surveillance has become a critical component in today's world. With developments in advanced systems have been developed as a result of the precision and efficacy of deep learning, machine learning, and artificial intelligence to identify and identify questionable activities while live image monitoring is taking place. Because human behavior is fundamentally unpredictable, it can be difficult to judge whether an action is suspicious or typical. Human activities and behaviors were divided into two categories in this study: suspicious and normal. Suspicious activities include sprinting, boxing, or fighting, whereas normal activities include sitting, strolling, jogging, and waving their hands. Convolutional neural networks are used to achieve this classification (CNNs). High-level features are first extracted from the photos by the CNN. The convolutional network's classification is then applied, and the final prediction is informed by the outcomes of the final pooling layer.