"Snap-n-Eat": Food Recognition and Nutrition Estimation on a Smartphone

Area of Science:

• Computer Vision
• Mobile Health
• Nutritional Informatics

Background:

• Accurate dietary assessment is crucial for health management.
• Previous food recognition systems often require controlled environments or manual user input.
• Automated tools for dietary intake estimation are needed for real-world applications.

Purpose of the Study:

• To develop and evaluate Snap-n-eat, a mobile system for automatic food recognition and nutritional content estimation.
• To enable users to easily track their food intake by simply taking a photo of their meal.
• To overcome limitations of prior systems by handling cluttered backgrounds and multiple food items without user intervention.

Main Methods:

• The system utilizes image processing techniques including salient region detection, background subtraction, and hierarchical segmentation.
• Feature extraction is performed at multiple scales and locations.
• A linear support vector machine classifier is employed for food item recognition.
• Portion size estimation is integrated for comprehensive nutritional analysis.

Main Results:

• The Snap-n-eat system demonstrates automatic food detection and recognition in real-life settings with cluttered backgrounds.
• It can simultaneously identify multiple food items and estimate their portion sizes.
• Experimental results show an accuracy exceeding 85% for detecting 15 different food types.
• The system was implemented as both an Android application and a web service.

Conclusions:

• Snap-n-eat offers a novel, automated solution for food recognition and nutritional analysis using mobile technology.
• The system's ability to function in unconstrained environments and handle complex meals represents a significant advancement.
• This technology has the potential to enhance dietary self-monitoring and promote healthier eating habits.