A Portable, Low-Cost Eye Tracking System for Predicting Visual Attention During Meal Ingestion

Obesity is a pressing public health challenge affecting physical and mental health. Understanding the cognitive and socio-economic factors that influence eating behaviors is essential for effective interventions. Visual attention is crucial in food choices, making it critical to study how individuals engage with food in real-life settings. However, current methodologies often rely on static images, limiting their ecological validity and real-world applicability, or are highly intrusive. This work offers a practical solution for tracking visual attention during real-life eating scenarios. We designed a tray with an integrated camera and a calibration pattern to remotely record volunteers during meal consumption. A convolutional neural network (CNN) was employed for the gaze tracking, utilizing a custom-built training database. Models without and with user calibration were provided. Results indicate an accuracy error of 5 cm, which improves to 2 cm with calibration, demonstrating our system’s effectiveness in measuring visual attention in food psychology experiments.