Intelligent Diagnosis and Prediction of Plant Diseases Based on Efficientnet_B3a

Abstract

Crop diseases pose a significant threat to agricultural production, affecting plant growth, reducing yields, and ultimately hampering the agricultural economy. Therefore, accurate detection of plant diseases is paramount. Recent advancements in deep learning have demonstrated superior performance compared to traditional recognition methods in the realm of intelligent diagnosis. Motivated by this, the present study employs a multi-network model based on transfer learning to identify 39 distinct types of plant diseases. We innovatively select the Efficientnet_b3a network model for plant leaf disease recognition. The Efficientnet_b3a is used for the first time in plant detection. This model boasts the advantages of high accuracy and low network complexity. To further validate its performance, we utilize three established network models, namely Resnet50, Google Net, and Inception_V3, as benchmarks for comparison. To validate the effectiveness of the Efficientnet_b3a model, this paper employs the publicly available Plant Village dataset to conduct training, validation, and testing on four different network models for comparative analysis. The experimental results demonstrate that the Efficientnet_b3a model achieves a remarkable recognition accuracy of 99.13% in plant disease identification, outperforming traditional network models. Notably, it manages to reduce the complexity of the network and improve the efficiency. At the same time, it is used in the rapid identification of some complex scenes. Compared with the classical complex network, the method verified in this paper has advantages in the rapid diagnosis of plant diseases. Furthermore, the model exhibits excellent network complexity, making it suitable for plant leaf disease recognition in practical agricultural production. This provides an efficient and accurate solution for disease detection in agricultural production, thus exhibiting broad application prospects.