Lightweight human ear recognition method and system based on YOLOv5s

The invention discloses a lightweight human ear recognition method and system based on YOLOv5s, and the method comprises the steps: improving a YOLOv5s algorithm architecture, employing a MobileNetV3 network as a backbone network of a human ear recognition model, and constructing the human ear recognition model; the method comprises the following steps: acquiring human ear data, performing human ear recognition through a constructed YOLOv5s-MG human ear recognition model, performing feature extraction through a MobileNetV3 network of the human ear recognition model, performing feature fusion through a neck network of the human ear recognition model, and performing human ear recognition through an output network. According to the method, a YOLOv5s algorithm architecture is improved, feature extraction of a backbone network and/or feature fusion of a neck network of YOLOv5s are simultaneously lightened, so that the parameter quantity, the calculation quantity and the model size of the network are greatly reduce