Review of Violence Detection and Alert System in Video using Deep learning
Tejas Bose, Omkar Gaikwad · INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT · 2025
The increasing prevalence of surveillance cameras for monitoring human activities necessitates automated systems capable of detecting violence and suspicious events. The detection of abnormal and violent actions has emerged as a significant area of research in computer vision and image processing, attracting considerable interest from researchers. This paper provides a comprehensive review of recent advancements in violence detection techniques. The methods reviewed are categorized based on their classification approaches, including traditional machine learning, Support Vector Machine (SVM)-based methods, and deep learning techniques. Additionally, this study highlights the feature extraction and object detection methods employed in each category. The datasets and video features that contribute significantly to the recognition process are also analyzed. To enhance understanding, an architectural diagram is presented to illustrate the key steps involved in the reviewed approaches. The findings of this review aim to guide future research by identifying gaps and opportunities for advancements in the field of violence detection.