Crop Pest and Diseases Classification Using ResNet and Inception Network

Traditional identification and classification methods of pests and diseases mainly rely on manual experience and feature extraction, which have the disadvantages of low accuracy and being easily affected by lighting, angle, and other conditions. Therefore, how to identify and predict crop pests and diseases timely and accurately has become an important issue in agricultural production. Residual neural network (ResNet) solves the problem of gradient disappearance and overfitting in deep neural network training by introducing Residual Block. Networks can transmit information directly across multiple levels, thus achieving a deeper network structure. In this study, an improved and optimized ResNet network was used to classify crop pests and diseases. First, the ResNet network is used to extract the high-level features of the image and map them into the output space. Then, the network is improved and optimized by combining inception structure, hyperparameter enhancement, and transfer learning to improve the performance of the model. The experimental results show that the optimized fusion network achieves better results on the crop pest image classification task.