Research on Apple Leaf Disease Identification Method Based on Improved Deep Residual Shrinkage Network
Download as PDF
Liu Pu, Wang Xun
In response to the problems of slow recognition speed and inaccurate recognition in traditional methods for identifying apple leaf diseases, this paper proposes an improved deep residual shrinkage network recognition method. This method takes the deep residual shrinkage network as the basic framework and introduces the Inception module to efficiently extract disease features at different scales, enhancing feature diversity. At the same time, double transfer learning is used to enhance the generalization ability of the model on the task of apple leaf disease with small samples, and reduce the impact of data on the model performance. The experimental results show that the proposed method achieves an accuracy of 98.8% in identifying apple leaf diseases. In addition, compared with traditional methods, the proposed method has significant improvements in many aspects such as convergence speed and robustness.
Deep residual shrinkage network; Apple leaf disease; Dual transfer learning; Multiscale convolution