Image-based food portion size estimation using a smartphone without a fiducial marker.

Area of Science:

• Bioengineering
• Dietary Assessment Technology
• Medical Imaging

Background:

• Current food volume estimation methods require inconvenient fiducial markers, leading to time-consuming and inaccurate post-processing.
• These limitations hinder the use of smartphones for self-administered dietary assessments.

Purpose of the Study:

• To present a novel smartphone-based imaging approach for table-side food volume estimation.
• To overcome the limitations of existing methods by eliminating the need for fiducial markers.

Main Methods:

• Developed a fiducial-marker-free food volume estimation method using smartphone imaging.
• Utilized a mathematical model incorporating smartphone physical length and motion sensor data for calibration.
• Introduced a virtual reality method with the International Food Unit™ and automated training for error control.

Main Results:

• A pilot study with 69 participants and 15 foods validated the fiducial-marker-free approach.
• The training process significantly improved food volume estimation accuracy (P<0.05).

Conclusions:

• The developed method enables quick food volume estimation without fiducial markers.
• Virtual reality, the International Food Unit™, and automated training enhance estimation accuracy and control errors.
• Estimated food volumes can be used for nutrient database searches to determine dietary energy and nutrient intake.