Men's Shirt Collar Pattern Recognition Based on YOLO11

This study addresses the issue of collar type recognition in the custom production of men's shirts and innovatively proposes a classification method based on the YOLO11 deep learning model. By integrating clothing structure parameters and sensory features, a dataset of 2200 shirts was constructed, containing six basic collar types: Point collar, standup collar, Button-down collar, Spread collar, Cuban collar, and Wingtip collar. To validate the accuracy of the YOLO11 model, experiments were conducted comparing three models: YOLO11, DenseNet201, and ResNet152. The results showed that the YOLO11 model achieved an overall accuracy rate of 98.0% in collar type classification, representing an improvement of 19.0% and 21.6% compared to the comparison models. and demonstrated outBand performance in fine-grained classifications such as button-down collars (95.8%) and Wingtip collars (98.2%). This study integrates object detection technology with clothing structural features, providing a high-precision recognition solution for shirt intelligent manufacturing. It holds significant application value in style recommendation and process decision-making for personalized custom production.