Fully Connected Deep Learning for Tomato Disease Detection and Classification using Grey Co-Occurrence Matrix and Haar

Tomato crops are vulnerable to a variety of diseases, which can result in severe financial losses for producers. Traditional methodologies for disease detection and classification are labor-intensive and necessitate specialized knowledge. In the recent years, advancements in deep learning models have shown favorable results in automating this process. In this study, we introduce a Fully Connected Neural Network deep learning architecture aimed at the identification and categorization of tomato diseases. The implementation of image segmentation via K-means clustering utilizing HSV color space is employed to enhance the quality of the outcomes. The suggested methodology employs the Grey Level Co-occurrence Matrix (GLCM) and Haar to facilitate the extraction of significant image features. It also compares the result of both the feature extraction techniques using Decision tree classifier, Random Forest, Principal Component Analysis and KNN. We utilize the tomato-leaf-multiple-sources dataset from Kaggle along with Real Image dataset to systematically train and assess the efficacy of our model. Our suggested model exhibits superior performance compared to existing state-of-theart techniques in the diagnosis and categorization of tomato diseases using Fully Connected Neural Network.