EAT: A Reliable Eating Assessment Technology for Free-living Individuals.

Summary

This project aims to develop and validate a reliable, objective eating assessment technology (EAT) for free-living individuals, utilizing machine learning with wearable cameras and IR sensors to overcome self-report biases and address privacy concerns through obfuscation techniques.

What they want

The project involves developing an activity detection algorithm using IR sensor array data and RGB images to detect eating. It will test various obfuscation methods in a cross-over trial to select the most participant-acceptable approach. The chosen algorithm and obfuscation method will be deployed on a novel wearable camera with an infrared sensor array to test eating detection in real-world settings. The algorithm's performance will be validated against real-time user responses and 24-hour dietary recall.

Deliverables

Activity detection algorithm (using IR sensor array and RGB images)
Selected best obfuscation method (based on participant acceptability)
Novel wearable camera (with infrared sensor array, deployed with algorithm and obfuscation)
Validated eating detection algorithm (performance compared against real-time user response and 24-hour dietary recall)

Technical requirements

Machine learning for detecting eating in videos
IR sensor array data processing
RGB image processing
Wearable camera technology
Cross-over trial methodology
Real-time user response for validation
24-hour dietary recall for validation