A System to Estimate the Amount and Calories of Food that Elderly People in the Hospital Consume

Malnutrition in the elderly is an important issue, especially for the elderly in hospitals. In general, nutritionists calculate energy from foods that the elderly need each day. If the elderly eat less food and do not receive enough energy, supplementary food must be provided. Currently, the problems in hospitals are the lack of nutritionists and the calculation of the amount and energy of food takes a long time. In this research, we present a system to estimate the amount and calories of the food that the elderly in the hospital consume by taking pictures in a divided food tray before and after consumption. Our system has three main parts: food detection and classification, food weight estimation, and calories estimation application. In the first part, we use the Faster R-CNN technique and select ResNet-50 as a pre-trained model. Our prediction model was trained on a Suandok Hospital Food Images Dataset (SH-FID) of 16, 067 food images from 39 different classes. The result with 4017 food images showed that mAP = 73.354. In the second part, we use the CNN technique, which uses a pre-trained model as InceptionResNetV2. From the experiment with the 4017 food images found that MAPE = 16.9729, which is considered a good prediction. Finally, we created a web application to display the number of calories consumed by the elderly by converting the calculated food weight into calories using the hospital reference table.