Novel Approaches for Video Analytics

Abstract

Video Analytics is a booming research area in modern days. It analyses the video newlinecontent and adds brains to the eyes i.e. analytics to the camera. It extracts contents from the newlinevideo by monitoring the video in real-time. Normal and Abnormal human activity detection newlineusing deep learning models is a significant topic in computer vision. It will help in reducing the newlineincreasing crime rate by preventing treacherous actions leading to public safety and security. newlineThere is a huge advancement in human action recognition and prediction-related research but newlinestill, state of art algorithms need improvement in classifying it. The proposed method focuses newlineon abnormal activity detection for humans. Hybrid approaches with customized cost functions newlineare used using deep neural networks. Transfer learning technique is implemented along with it newlinewhich helped to improve efficiency in terms of time complexity. Also, a video dataset is created newlinefor Indian crime scenarios using YouTube videos for the above system. The proposed method newlineshows improvement in accuracy as compared to the previous state of art methods. newline