Design of a Multi-Object Wearable Recognition and Tracking Algorithm Based on PHSM-YOLO

The objective of this project is to design and develop a multi-object wearable recognition system using advanced deep learning-based object detection and classification models. The project mainly focuses on detecting important safety-related wearable objects and human presence, such as Helmet, Vest, Person, bare-arms, Gloves, Non-Helmet, and Shoes. The main aim is to identify whether a person is wearing the required protective equipment in workplace, construction, and industrial environments. To achieve this, two advanced models are implemented: YOLOv8+BiFPN+SimAM+DyHead and RT-DETR+CST+CBAM. The YOLOv8-based model improves object detection through enhanced feature fusion, attention learning, and dynamic detection head mechanisms. The RT-DETR-based model improves detection accuracy by combining transformer-based object detection with attention modules. The project also includes both classification and object detection, allowing the system to classify wearable categories and locate them using bounding boxes. The final objective is to build an automated, accurate, and efficient system that can support safety monitoring and reduce manual checking.