Transfer Learning for the Identification and Application of Foliar Diseases in Agricultural Crops

Foliar diseases of agricultural crops seriously affect the yield and quality of crops, but the number of corresponding crop foliar disease image samples is limited, which leads to training difficulties. In order to solve this problem, this paper innovatively proposes a transfer learning-based method that uses the pre-trained model trained on a large-scale dataset for feature extraction and parameter adjustment, and transfers the knowledge on the large-scale dataset to a small-sample dataset for training and classification of cassava leaf disease. The experimental results confirmed that the method effectively overcame the problem of small samples, significantly improved the identification accuracy of cassava leaf diseases, and provided a powerful tool for the diagnosis and analysis of cassava leaf diseases.