Enhancing Crop Health and Early Detection of Tomato Leaf Diseases Using Deep Learning Techniques

Abstract Background/Objectives: One of the most widely produced and consumed crops in the world, tomatoes are often threatened by various leaf diseases, including early blight, late blight, and leaf mold, which can result in significant yield losses if not detected and managed promptly. Due to reliance on manual inspection, tomato leaf diseases are often detected late, resulting in significant crop loss. The goal of this research is to develop a deep learning (DL) model that accurately classifies tomato leaf diseases. The model is trained using supervised learning on the PlantVillage dataset, which includes labelled images of tomato leaves under different conditions. Method: Gaussian Blurring-Gaussian Mixture Models (GB-GMM) are a preprocessing method used to enhance the image quality. While EfficientNet and VGGNet architectures are utilised for accurate classification, data augmentation is used to boost model robustness. A standard train-validation-test split is used to assess the model. Findings: The results demonstrate that the proposed method, implemented in Python, performs better than the others in terms of accuracy (94.3%), precision (93.7%), recall (92.5%), and F1-score (93.1%) in VGGNet architectures. These results indicate that the proposed model is very effective overall and produces balanced predictions. In the future, the system may be integrated with drone and IoT technologies for automatic disease warnings and real-time field surveillance. Novelty: The Multivariable Grey Prediction Evolution Algorithm (MGPEA) is included for illness trend forecasting to enhance predictive power further. This technology facilitates large-scale, sustainable agricultural management, minimizes human inspection, and enables prompt disease response. Keywords: Tomato Leaf Diseases, Crop Health, Gaussian Mixture Models, Multivariable Grey Prediction Evolution Algorithm, EfficientNet, VGGNet