Optimized Tomato Leaf Disease Diagnosis Using Feature Extraction and Inception-ResNet-V2

Tomato leaf disease poses a significant challenge as it can reduce crop yield and degrade fruit quality. To facilitate automatic and precise detection, this study leverages a deep learning approach combined with feature extraction techniques, namely Gray Level Co-occurrence Matrix (GLCM) for texture analysis and Hue, Saturation, Value (HSV) for color representation. The dataset utilized in this research comprises 17,146 tomato leaf images affected by various diseases. The classification process employs the Inception-ResNet-V2 architecture, a convolutional neural network known for its superior performance in image classification tasks. Initially, GLCM is applied to extract texture characteristics, while HSV is used to capture detailed color information. These extracted features are subsequently fed into the Inception-ResNet-V2 model, which is trained on the processed dataset. Additionally, hyperparameter tuning is conducted to optimize model performance. Experimental results indicate that integrating GLCM and HSV features with the Inception-ResNet-V2 model yields an accuracy of 98.65%. The training process is conducted using the Adam optimizer over 10 epochs, with data split into 70% for training, 20% for validation, and 10% for testing. The findings demonstrate that this approach is highly effective in accurately predicting tomato leaf diseases, making it a viable early detection system. By implementing this method, early intervention strategies can be enhanced to curb the spread of diseases in tomato plants, thereby improving agricultural productivity.