Research on Zero-Shot Automatic Segmentation of Safety Harness Based on CLIP

This article aims to explore computer vision research on safety belt recognition in electric power tower construction sites. Proper wearing of safety belts is crucial for the safety of workers during tower construction. However, most existing methods for safety belt recognition often rely on manual annotation, resulting in low efficiency and potential errors. Therefore, the use of Zero-Shot Learning (ZSL) for automated safety belt recognition is of great significance. This study proposes a CLIP-based image recognition technique for the automatic identification of whether construction personnel are wearing safety belts at tower construction sites. This method is characterized by automation, efficiency, and accuracy, providing important safety assurance for tower construction. Future work can further explore safety belt recognition in complex backgrounds and expand applications such as joint detection with other safety equipment.