 The proposed method is a lightweight ear recognition method based on an attention mechanism and feature fusion. It has improved the accuracy of the YOLOV5S-MG method by adding the CBAM attention mechanism and SPPF layer, as well as cross-regional feature fusion. Furthermore, the proposed method has achieved a higher map at 0.5 score compared to the YOLOV5S-MG method and the YOLOV5S method. Additionally, the proposed method has satisfied the requirements of real-time performance and high accuracy. This article was authored by Yen Min Le, Dong Pan, Ji Bing Feng, and others.