Smart automated fruit freshness recognition system using image processing and deep learning

Abstract

Bangladesh is one such country with a tropical monsoon climate typified by significant seasonal rainfall, high temperatures, and high humidity. A wide range of tropical and subtropical fruits are abundant in Bangladesh. The fruits that are most frequently grown are mango, jackfruit, pineapple, banana, litchi, lemon, guava, wood apple, papaya, tamarind, watermelon, pomegranate, plum, etc. Automated fruit recognition is essential since fruits in Bangladesh’s markets come in a variety of types and qualities. This thesis presents a deep learning-based automated fruit recognition model that uses image processing and deep learning architecture to identify fruits and grade their quality. We will make use of our dataset of Bangladeshi fruits for the experimental evaluation. This thesis aims to provide a novel Convolution Neural Network (CNN) structure, called VGG19, for identifying, classifying, and evaluating fruit objects according to their freshness. An application for Keras called VGG19 has a high degree of accuracy in object detection. The outcomes show that our method works better than the linear predictive model and demonstrate its particular merit.