Plant Leaf Classification and Recognition Based on Deep Learning

In order to address the problems of insufficient edge feature information extraction and large leaf positioning deviation caused by complex background of leaf images, an improved plant leaf classification and recognition algorithm based on YOLOv5 was proposed in this paper by introducing attention mechanism and optimizing loss function. Adding attention mechanism can enhance the correlation between target features in location and channel, which is beneficial for the network to extract and utilize key feature information of leaf edges better, thereby improving the accuracy of network recognition. Adjusting and optimizing the loss function can reduce the localization error of prediction boxes and improve the recognition rate of leaf in dense areas. The experimental results show that the mAP of improved YOLOv5s exceeds 94%, which is 3%-6% higher than other models, effectively enhancing the classification and recognition performance of the original detection model.