Agriculture is a primary industry for sustainability and growth of humanity. High crop yielding is the basic requirement to feed the current population of this globe. Plants has a vital role to play in biodiversity sustenance. Precision farming or precision agriculture is the practice to maximize the crop yields and make agricultural profession more profitable. Precise and timely input of various agricultural parameters through smart and advanced technologies like IoT (Internet of Things), AI (Artificial Intelligence), Image Processing, Computer Vision, Drone based cameras, smart portable devices, GPS and others are providing precision farming a real playground for implementation. The practice of precision farming can boost the efficiency, sustainability, and profitability of farmlands. The vegetables and fruits plants not only demanded in agricultural productivity, but also in manufacturing of medical products, Cosmetics products, herbal and organic products and many more. Tomato is one major food crop in agricultural crops across the globe. There is approximately 20 kilogram per capita consumption per year of tomato and it represents 15% share in average total consumption of vegetables. To meet a huge demand of tomato worldwide, it is required to develop new techniques or improvise the existing ones for improving crop yield and early detection of diseases caused by viral infection, pests or bacteria. Early detection of such disease in tomato plant will help to increase productivity and quality of tomato. Convolutional Neural Network based models for recognizing and classifying of tomato leaves disease is proposed in this paper. Total 22930 images of tomato leaves (healthy and un-healthy) are used to train and validate the proposed CNN Based model and acquired an accuracy of 98.7%. ROC (Receiver Operating Characteristic) curve and other statistical parameters, including specificity, sensitivity, recall, precision and accuracy was applied to compare the performance of various pre-trained model used in transfer learning. The results are clearly indicates that AUC (Area Under Curve) values for implemented models are high (AUC>0.92).