A Lightweight Factory Unsafe Behavior Recognition Method With Occlusion-Robust Adaptive Fusion

This paper presents a novel approach to real-time factory unsafe behavior recognition using YOLOv8-CBAM, RT-DETR-Swin, CBAM, and Adaptive Fusion. The system aims to identify unsafe behaviors, specifically smoking behavior, by processing image datasets through advanced object detection techniques. The YOLOv8-CBAM and RT-DETR-Swin models are employed for detecting objects in images, with Adaptive Fusion applied to enhance the accuracy of object localization by reducing redundant predictions. CBAM is utilized for visual interpretability, providing insights into the decision-making process of the models by generating class activation maps. The detection process focuses on a single class: smoking, and optimizes the performance by fine-tuning hyperparameters and leveraging the strengths of these advanced models. The methodology improves the robustness and precision of unsafe behavior recognition in real-time applications, offering a scalable solution for automated factory safety and compliance systems.

Keywords: YOLOv8-CBAM, RT-DETR-Swin, CBAM, Adaptive Fusion, factory unsafe behavior recognition, smoking behavior detection, object detection, hyperparameter optimization, machine learning, deep learning.