The cutting edge surface of precision tools with raised teeth is prone to wear and impact damage during high-frequency use.These defects are characterized by their small size and diverse textural appearances,which pose significant challenges for traditional machine learning models to achieve high-precision detection.To address this issue,this paper proposes a tool defect detection method named CutterNet,based on an improved YOLO-v8(you only look once version 8) framework.Firstly,a cross-stage partial multi-scale attention module (CSPMSAM) is introduced to extract local defect features of the tool, thereby enhancing the model′s capability to detect defects at various scales.Secondly,the asymptotic feature pyramid network (AFPN) is incorporated to improve the fusion of multi-scale features and reduce the semantic gap among them.Finally,the original complete intersection over union(CIoU) loss function in YOLOv8 is replaced with Inner-CIoU to enhance the regression accuracy of bounding boxes. Experimental results demonstrate that the improved algorithm achieves a 3.1% increase in detection accuracy,a 30.12% reduction in model parameters,and an improvement in inference speed from 58 to 60 FPS.These performance metrics surpass those of most other mainstream object detection models.The proposed algorithm has been successfully implemented in a real-time tool defect detection system.