Android-Based Application for Ripeness Classification and Shelf-Life Determination of Philippine Bananas

Lauded as the world's most popular fruit, bananas are one of the main exports of the Philippines. It contains essential nutrients that can have a protective impact on health. Commonly, farmers can distinguish the different types of bananas and potentially predict their shelf-life. Nevertheless, human judgment is not always reliable. It needs to have support and evidence to minimize the margin of error. Hence, this study focuses on developing an android application to determine the Philippine bananas' ripeness, variety, and shelf-life. This application identifies different bananas sold in the local market, specifically Cardava, Lakatan, and Latundan. Also, the application determines the shelf-life of bananas, such as the remaining days for unripe bananas to become ripe and ripe bananas to overripe. A Convolution Neural Network-based (CNN) model classifier is used as an architecture categorizing the Philippine local market bananas. The proponents used real-life image data to increase the accuracy of the CNN model. The application was proven effective and has reached 90% accuracy.