Machine Learning Approaches to Classify Indian Mango Varieties

Mango is one of the most common tropical fruits and is known as king of fruits. It is the most loved fruit in India. Currently, the classification of mangoes is still primarily done through manual inspection, which makes it difficult to automate, time-consuming, and unreliable. So, the classification of mango fruit using image processing, computer vision, and artificial intelligence methods is needed. This methodology helps the mango industry in increasing productivity, profitability, and export potential. In this study, we introduce an effective machine vision system for better classification which can reduce the wastage and at the same time increase the marketing of mango export. In this paper, the mangoes are classified using a convolutional neural network (CNN). In this study, we classify mango images into 10 categories: Alphonso, Banganapalli, Chausa, Dasheri, Kesar, Malgova, Mallika, Neelam, Raspuri, and Totapuri. Natural image preprocessing is essential to image processing since unprocessed natural images might lead to inaccurate classification if they are fed straight into the CNN model. Thus, an improper classification can result. In the experiment section, the natural photos of mangoes were first preprocessed, and the convolution neural network was then fed with these preprocessed images as input for the classification.