Deep CNNS and fractal-based sequence learning for violence detection in surveillance videos

Deep learning has received a great deal of attention from researchers for its reasonable outcomes in different fields of computer vision due to their wide range of applications. The recognition of different activities in videos is important for violence detection (VD) to ensure safety and security. VD also practice the latest deep learning-based algorithms. This chapter provides a detailed overview of CNN-based state-of-the-art VD methods and visual representation of their working strategy. Next, we also discuss the concept of sequence learning and its different variants for violent scene detection by investigating their internal mechanism. Similarly, we provide comprehensive detail of comparing the VD methods considering the accuracy obtained on different challenging datasets. Furthermore, the concrete details pertaining to the performance of the methods using standard metrics is provided. Finally, we discuss the main necessities of VD methods and future research directions.