Automated Artificial Intelligence-Based Thai Food Dietary Assessment System: Development and Validation

Area of Science:

• Nutrition Science
• Artificial Intelligence
• Computational Biology

Background:

• Traditional dietary assessment methods for Thai cuisine are labor-intensive and prone to errors.
• Accurate dietary assessment is crucial for nutrition research and chronic disease management.

Purpose of the Study:

• To introduce INMU iFood, an AI-powered system for estimating the nutritional value of Thai dishes from images.
• To enhance dietary assessment accuracy and efficiency in the context of Thai food.

Main Methods:

• Utilized an AI-based system, INMU iFood, employing image recognition, manual input, and barcode scanning.
• Upgraded the core model from Yolov4-tiny to Yolov7 and expanded the training dataset with non-carbohydrate foods.
• Integrated with Open Food Facts database via barcode scanning for access to over 3000 food items.

Main Results:

• The Yolov7-based INMU iFood system demonstrated improved accuracy in food item identification, particularly for complex images.
• Significant enhancements were observed in protein and fat estimation compared to the previous Yolov4-based system.
• Barcode scanning integration expanded the system's utility, providing access to a comprehensive nutritional database.

Conclusions:

• INMU iFood offers a promising, accurate, and versatile tool for dietary assessment of Thai cuisine.
• The system supports researchers, healthcare professionals, and individuals in monitoring dietary intake and improving health outcomes.
• INMU iFood facilitates nutrition research and promotes better health through enhanced dietary tracking.