Fruit-Classification

Project Description

Independent living is considered an essential part of society, but activities of daily life, such as food preparation, come with additional challenges for people with vision loss. While an increasing number of advanced vision assistive devices are entering the market, assisting with food safety, such as food ripeness or moldiness, has received little attention. For this project, a model was developed to assess the quality of fruit from an existing data set, which could be integrated into a product for use in home kitchens. Four fruits (banana, apple, orange, lime) were classified with four labels (unripe, ripe, overripe, moldy) using three machine learning mod- els. The accuracy of all three models was assessed: convolutional neural networks (87.31%), ResNet50 (70%), and ResNet50 (92.72%). The precision, accuracy, and F1 score was also evaluated for all models, and opportunities to improve on the model and make the final product more viable are discussed.