An Intelligent Approach to Determine Banana Ripeness Stages using Deep Learning Models

Fruit ripeness classification is important in agriculture, food processing industry, supermarkets, Ayurveda, and many more. The manual identification is tedious, time-consuming, and laborious. Hence, deep learning techniques have gained popularity recently as a computer vision method for automatically classifying fruit ripeness simplifies feature engineering work. Differentiating between the visual characteristics of fruits at different stages of ripeness is a difficult task for machine learning systems. In this proposed work, two CNN models such as InceptionV3 and VGG19 are used to invent the best deep-learning model to categorize bananas into their three distinct ripeness stages. The experiments are carried out using self-created datasets consisting of 687 image samples of three different stages. The first dataset consists of individual banana image samples and the second dataset contains banana bunch sample images. The CNN models are trained and tested in the range of 80:20. It is observed that VGG19 achieves better accuracy than InceptionV3 in predicting three ripening stages of banana fruit with an accuracy of 98.1% and 97.5%.