The crops 'disease detection lies at the heart of agricultural sustainability, and tomatoes-one of humanity's most precious resources-are afflicted by many leaf diseases. The idea of this research is the presentation of a new method based on Convolutional Neural Networks (CNN) for detecting tomato plant leaf disease. Its objective is to improve accuracy by utilizing a segmentation strategy and feature extraction models. This methodology includes the collection of a number of public and private data sets as well as the development of an additional regional-specific dataset for Chhattisgarh in India. This segmentation uses the mean-shift image segmentation strategy which more effectively isolates lesions than traditional techniques. Extraction of shape, texture and colour features used Principal Component Analysis (PCA) to reduce the dimensionality. Pretrained functions, such as VGG-16 go one step further in selecting which features will be used for classification. Phase three is classification. Usually, the classifier goes through machine learning, which allows the selection of models such as SVMs (support vector machines) or decision trees. The results should be a region-specific set of parameters, segmentation which is most suitable for the task and an accurate classifier model. This research builds on past work involving tomato disease detection and extends those advances by uniting segmentation, feature extraction, and classification in a single CNN. Its importance lies in giving farmers an automatic disease alarm system that is also economical, promoting sustainable agriculture.