Prediction of tomato leaf disease using deep learning approach

Abstract

Abstract Diseases of tomato leaves are significant threats to the global food security and agricultural production. The old method of diagnosis is not reliable and is time consuming, and there is a demand to have effective and accurate automated systems. The paper uses transfer learning using Inception-V3 and Inception-ResNet-V2 network to detect tomato leaf diseases using an open dataset. To encourage generalizability, data augmentation and preprocessing techniques were used, whereas Grad-CAM was used to encourage visual interpretability. Experimentally, it has been demonstrated that Inception-ResNet-V2 and Inception-V3 performed with 92.33 and 89.33 accuracy, respectively, which is higher than the other existing methods. These results demonstrate the possibility of deep learning to improve precision agriculture and prepare further development of real-time and field-deployable systems of disease detection.

Code and data availability

The paper's Data Availability Statement explicitly identifies the public Kaggle tomato leaf image dataset used to train and evaluate the deep learning models, making it a paper-specific, publicly actionable asset. No author analysis code or trained model checkpoints are disclosed.