Analysis of Tomato Leaf Diseases using a Deep Learning Model

---

Abstract

Plant leaf diseases cause substantial annual production losses for farmers, impacting their primary food source. Minimizing these losses requires early detection of diseases. A deep learning approach is used to address the early identification of tomato leaf diseases, enabling farmers to take preventive measures and reduce production loss. A Customized six-layer Convolutional Neural Network (CNN) is suggested for the identification of diseases in tomato plant leaves, aiming to mitigate annual production losses for farmers. The CNN utilizes automatic feature extraction, requiring no explicit feature engineering, enabling finer disease classification. Employing 11,100 leaf images from the Plant Village dataset, the model classifies ten classes, including nine distinct diseases and one healthy class. With an 80/20 dataset split, 30 epochs, and a 0.001 learning rate, the CNN achieves a overall accuracy of 91.3%. The focus on computational efficiency addresses the critical issue of early and accurate disease detection, potentially boosting agricultural productivity and affordability for consumers. Comparative analysis with VGG-16 and VGG-19 using transfer learning reveals the superior performance of the proposed model. Its simplicity not only reduces parameters but also facilitates deployment on lightweight devices, significantly reducing training time and the simulation utilized Google Colab. It also emphasizes the effectiveness of a simplified approach in addressing crucial challenges in tomato disease identification, with potential applicability to other crops and plants.

---

Keywords

Deep learning, Customized CNN, Transfer learning, Leaf disease classification, hyperparameters.