Improved Weld Inspection Model of Ship Pipe System Based on YOLOv5

Ship tube system plays a crucial role in the design, construction, operation and maintenance of ships, which will have an impact on the service life of the ship, and requires strict control of the quality of the tube system, so high accuracy is needed in the inspection of the weld seams of the tube system. In this paper, an in-depth improvement of the YOLOv5 target inspection model is carried out around the characteristics of weld defects in the pipe system, with the goal of enhancing the model's ability to capture the characteristics of weld defects. To address the challenge of the small size of some defects in the weld, the model was extended to include a 160*160 detection layer specialized for small targets to capture fine features in a more refined manner. Integration of the CBAM attention mechanism into the YOLOv5m model enhances the focus on critical features, which significantly improves the detection accuracy of the defective region. The original YOLOv5 model as well as the improved YOLOv5_CBAM model were subsequently trained and ablation experiments were implemented to verify the effectiveness of the improvements. The improved YOLOv5_CBAM model achieves 90.9% in mAP (mean average precision), which is 6.5% better than the original model.