Segmented inspection method for workwear based on human keypoints

In corporate environments, the automated detection of mixed-and-matched workwear faces significant challenges due to employees' diverse styles and inconsistent wearing habits, leading to high misidentification rates with traditional methods. This study introduces a novel segmented detection algorithm based on human body keypoint estimation technology. Using the YOLO-Pose model, human silhouettes and critical workwear zones are meticulously localized, isolating the upper and lower body regions. A specialized workwear recognition system, developed using deep learning methodologies with the YOLOv8 model, accurately distinguishes between summer and winter workwear, categorizing outcomes into three attire statuses. This advanced algorithm significantly improves the recognition of upper and lower garments, ensuring accurate identification even when garments are mixed and matched, reflecting actual workplace scenarios. The framework enhances the evaluation of employee adherence to dress codes and provides robust data-driven insights for intelligent management strategies and informed decision-making.