Image-Based Volume Estimation for Food in a Bowl.

Area of Science:

• Nutrition science
• Computer vision
• Food science

Background:

• Image-assisted dietary assessment is increasingly used for monitoring.
• Estimating food volume from 2D images is challenging due to lack of 3D data and occlusion.
• Accurate food volume estimation is crucial for precise dietary intake analysis.

Purpose of the Study:

• To develop a method for accurate food volume estimation in bowls using 2D images.
• To investigate the relationship between observable food surface and bowl fullness.
• To establish a reliable food volume assessment technique for dietary studies.

Main Methods:

• A mathematical model was created to describe various bowl shapes.
• An experimental method was proposed for determining bowl geometry.
• A Food Area Ratio (FAR) was developed to quantify food volume from images.

Main Results:

• A strong correlation was found between bowl fullness and the calculated FAR.
• Simulations and real-world experiments validated the proposed approach.
• The FAR method demonstrated feasibility and accuracy in estimating food volume.

Conclusions:

• The proposed FAR method offers a viable solution for estimating food volume in bowls.
• This technique can enhance the accuracy of image-assisted dietary assessments.
• The approach addresses limitations of 2D imaging in food volume quantification.