Food Recognition and Calorie Measurement using Image Processing and Convolutional Neural Network

The ease with which food is being delivered at our doorsteps has lead to an outbreak of a major chronic disease known as obesity. As the necessity of the food arose among people, the apprehension related to their diet also simultaneously increased. In this paper we propose a calorie measurement system whereby the user is made to upload the image of food item and as a result, number of calories present in the uploaded food image will be predicted. It is a multi-task system which also displays the weekly statistics on how much calorie is consumed by the user and how more/less calories must be consumed to avoid obesity related diseases such as heart attack, cancer etc. We built a dataset of food images collected from existing datasets to detect complex images consisting of 20 classes and each class containing 500 images each. We have curated our own Convolutional Neural Network architecture of 6 layers to extract the features and classify the images. Our experimental results on food recognition showed 78.7% testing accuracy with 93.29% training accuracy.