Predominantly, the identification of nutrient deficiency in plantain trees are based on visual symptoms. Nutrient anomalies are performed by trained specialists manually. It is a time-consuming process and needs proper attention since the symptoms appear in various parts of the crops. This work proposes an image classification and recognition system that analyses for analyzing nutrient deficiency from their visual appearance. The methodology uses deep learning techniques to analyze the nutritional status of the plant using the digital images of its leaves. Initially, the experiments are carried out for the boron and potassium nutrients deficiency of the plantain trees. The banana nutrient deficiency dataset consists of nutrient-deficient leaf images of plantain trees collected from various cultivation fields of Tamil Nadu. The efficacy of Convolutional Neural Network (CNN) makes them suitable for these kinds of applications. The system has experimented with state-of-the-art CNN architectures such as resnet50 and googlenet for classifying Boron and Potassium deficient plant images. The efficient architecture in terms of classification accuracy is selected for developing the SmartPhone-Centric real time nutrient deficiency detection system for assisting the farmers. The application presents the nutritional status of the plants, when getting the Plantain leaf images through the classification performed by the CNN. In the future, the dataset will expand by collecting the images of other nutrient deficient plants and will improve the classification accuracy with a greater number of images for obtaining more advantageous and expressive results for the real-time use of the farmers.