Food Recognition and calorie content measurement Android application.

The main purpose of this project is to recognize a food item from the pictures clicked by the user and then calculate its calorie content. This was my final graduation project and we worked in a group of four.

I developed an Android application with which the user has to click two pictures, a side view and a top view which gave the data to estimate volume of food item which in the next step helped to find mass of the food item and thus, the calorie content.

We used the dataset FOOD101 for training and used inception model in Tensorflow framework for recognition purpose. The images clicked by the user were sent to the local server and it recognized the food item within seconds with an accuracy of 84%.

For improving accuracy percentage, I displayed the top five suggestions generated by the machine to the user and asked them to select the best option so that the machine can be trained simultaneously. Then the approximate calorie content of the respective food item is displayed.

I also created a database of other nutrients like carbohydrates, proteins and fat and a database of recipes which was displayed with the calorie content. We are also working on adding a recommendation system so that the user can get food suggestions based on his/her previous searches.