Dangerous Object Detection Using YOLOv8 and Dynamic Snake Convolution
Biao Sun, Zhizhao Duan, Xun Han, Xiaobing Huang, Bincheng Xie, Xinchun Wu · 2024
This article studies the problem of object detection and recognition based on deep learning algorithms, focusing on the detection of dangerous items that can harm people in real life, heinous behaviors in fights, dangerous individuals such as masked criminals, and dangerous scenes such as fires. Deep learning object detection technology is of great significance for public safety. Through the analysis of video content, deep learning algorithms can effectively detect abnormal information, predict risks, and build intelligent monitoring systems. This not only improves monitoring efficiency, but also enables early detection and prevention of potential safety hazards, which plays an important role in maintaining social security and stability. In the experimental design, this article mainly focuses on the introduction of algorithm models, dataset production, model training, and recognition effects. Dynamic snake convolution has been added to YOLOv8 to enhance the detection performance of slender objects and improve the detection performance.