Mobile Application for Controlling a Healthy Diet in Peru Using Image Recognition

Abstract

Overweight is one of the big ills that affects the world population, especially the Peruvian population, and this is caused mainly by peoples ignorance of the amounts and nutritional values to consume according to their current condition. For this, there are various solutions focused on controlling a healthy diet for people, among the best known are the mobile food control applications such as MyFitnessPal, Fat Secret and MyNetDiary. These apps are quite useful for monitoring peoples food intake, as their databases have lots of food nutrition information. However, most of the information they have is focused on a foreign public, which may have different eating habits than Peruvians. That is why we present the application NutriCAM, which monitors the consumption of meals by users and provides the functionality of image recognition for meals, for the user to have a more friendly way to record and monitor its consumption, mainly focused on Peruvian gastronomy. The results are a Peruvian food recognition model based on the training of the pre-trained Convolutional Neural Network ResNet-50 and a dataset of 3600 food images, and a mobile application focused on the control of nutrition that caused 70% of improvement or maintenance in the current condition of 10 users.

Keywords

• mobile app
• image recognition
• food classification
• transfer learning
• peruvian food