Food Volume Prediction and Classification using Multi-Task Learning

Food volume prediction is critical for calculating calories in food analysis and nutritional evaluation. However, obtaining data and measuring volume from the 3D sensor camera concurrently is time-consuming. Additionally, high computing power is necessary to compute the volume. Therefore, machine learning techniques are employed to extract specific data collected from the data collection process for learning purposes to save time. In this study, we employ Multi-Task Learning to predict volume and categorize food due to their interdependence simultaneously. We evaluate several backbones and find that ResNet50 attains the best performance with an MAE of 4.87 for volume prediction and 100% food classification accuracy.