Pixel Bites-Estimation of Food Calorie For Diabetes Patients Using Deep Learning And Computer Vision

Abstract

One of the cutting edge fields right now are deep learning and computer vision. Using this we develop a mobile
app that takes a snap shot of the food we eat daily and recognize the amount of calorie and nutritional value present in the
food. It also posts a label whether the food is recommended for a diabetes patient or not. Because of this application, the
patient knows whether he can consume the food or not. Thus prevents the patient from various health issues. A good dietary
monitoring is also maintained by referring huge amount of data sets, since there are lots of different kinds of foods available
in the world. The biggest advantage is that this application does not require the usage of internet and hence can be used
anywhere at any time. Food recognition is based on the color, texture and size using various algorithms. Our system can
perform automatic food detection and recognition in real-life settings. These settings include cluttered background images.
Our system will also be able to identify multiple varieties of food present on a plate and simultaneously estimate their portion
size.