Deep learning solution for real-time violence detection in video streams
Sachin Tyagi, Shivansh Tyagi, Varun Bagga, Sourish Bansal, Siddharth Goswami, Yash Verma, Mohit Tyagi, SATYA PRAKASH SINGH · 2024
This paper explores the feasibility of detecting acts of violence in videos using advanced deep learning and AI techniques. The goal is to provide a straightforward implementation leveraging continuous advancements in technology. The paper emphasizes the importance of violence detection for safeguarding children from inappropriate content and helping guardians make informed viewing choices. The complexity lies in the abstract nature of violence, posing both technical and theoretical challenges. The study focuses on the application of pre-trained convolutional neural networks (CNNs) for a more accessible representation of violence. Transfer learning, a rapid and simplified approach, is employed with pre-trained networks designed for image categorization. The investigation explores various deep learning techniques to identify the most accurate model for violence recognition in video content.