Automatic weapon detection using Deep Learning
P Akshaya, Pamulaparthy Bharath Reddy, Pratheek Panuganti, P. Gurusai, Abdus Subhahan · 2023
Safety and security is important in today’s world scenario. For a country to be strong in terms of economy it must have secure and safe environment. CCTV surveillance is majorly used to monitor activities such as robbers and threats in public places, but these cameras need constant human observation. The fast and accurate automatic weapon detection system is useful to avoid these types of risks in public places such as schools, hospitals, museum and traffic etc. In many places the crimes are caused by pistols, guns and knives are very more, mainly in the countries and cities where there is no gun control laws. This work, focuses on automatic weapon detection in CCTV footage by making use of deep learning algorithms. In our work we are going to use YOLOv8 model to detect weapons as it is better in terms of accuracy and speed when compared to other models. It uses larger data set which consists of various images of weapons. Data set includes pictures from youtube robbery videos, CCTV footages, GitHub repositories and Internet movies firearms data base. Sliding window technique is used to extract features from objects and classify weapons. YOLOv8 has shown improvement in detection with greater speed and high accuracy.