Bolt position detection network based on key points algorithm

When using specialized robots for inspections of bolts on high-voltage transmission towers, an important task is to accurately locate the position of bolts in images captured by cameras, so that the mechanical device can proceed with further interactions. For this scenario, we use the YOLOv8 Pose model as the foundation and have incorporated a special attention mechanism to build a network that detects key points of bolts, while trying not to increase computational complexity. In our manually constructed dataset, the network we designed has demonstrated excellent performance, providing a solid technical foundation for further tasks.