YOLO-PTSW: an improved YOLOv8n-based algorithm for small target defect detection on swash plates

Abstract The swashplate is a critical component in automotive air-conditioning compressors, and its surface defects can directly affect the cooling performance and service life of the compressor. To address the challenge of accurately detecting small surface defects on the swashplate, we propose an improved YOLOv8n-based detection algorithm, named YOLO-PTSW. This algorithm incorporates a P2 small-target detection head to enhance fine-grained feature perception, introduces a 2wayConv downsampling module to improve information retention in shallow features, integrates the parameter-free simple attention module (SimAM) attention mechanism to strengthen feature responses, replaces the localization loss with Wise-PIoU to increase sensitivity to small-target boundary errors and improve localization accuracy, and applies pruning to achieve model lightweighting. Experimental results demonstrate that YOLO-PTSW outperforms the baseline YOLOv8n, with increases of 11.1%, 8.6%, 11.2%, and 17.5% in precision, recall, mAP@0.5, and mAP@0.5:0.95, respectively, while reducing the number of model parameters by 58%. These improvements not only enhance detection performance but also reduce model size, thereby improving deployment flexibility. Compared with other mainstream object detection models, YOLO-PTSW exhibits superior accuracy and efficiency, highlighting its strong potential for industrial applications.