Various diseases attack plant and lower the crop production rate across many countries. A few of the tomato diseases, such as Septoria Leaf Spot, Bacterial Spot, Late Blight, Early Blight, etc., cause heavy damage to the standing crop and reduce their production. If the diseases are detected at an early stage, then the agricultural losses can be significantly reduced. We propose a Convolutional Neural Network (CNN) based on automatic disease detection for tomato plants. The proposed CNN framework has four layers, and each layer performs a specific function. We were successfully able to classify nine tomato diseases using only the leaf images of the plant. This is preliminary research to explore the field of disease detection using neural networks. The neural network is trained with over 2000 images to form a model. The proposed model is validated on a standard dataset and yields substantial improvement in the state-of-the-art algorithms.