Fruit and Vegetable Information System Using Embedded Convolutional Neural Networks

This article presents the development of a mobile application that exploits a Convolutional Neural Network (CNN) to recognize a set of fruits and vegetables by processing snapshots taken by the built-in camera of the device. We built an acquisition system to gather pictures of different kinds of fruits and vegetables to train a neural network model. Instead of defining a new topology and training it from scratch, we took advantage of transfer learning and fine-tuned several MobileNet models to classify our images in their corresponding classes on a smartphone. Once the fruit or vegetable is identified, our mobile application provides valuable nutritional information about it.