Mobile Computer Vision-Based Applications for Food Recognition and Volume and Calorific Estimation: A Systematic...

Area of Science:

• Computer Vision
• Mobile Health (mHealth)
• Dietary Assessment

Background:

• Growing awareness of diet's impact on health drives demand for food analysis systems.
• Mobile applications are well-suited for real-time food recognition, volume, and calorific estimation.
• These systems aim to monitor consumption and support lifestyle changes for improved health.

Purpose of the Study:

• To systematically review mobile computer vision-based solutions for food recognition, volume, and calorific estimation.
• To assess the extent to which these applications provide explanatory feedback to users.
• To evaluate the potential of these applications in healthcare, particularly for chronic disease management.

Main Methods:

• Systematic review of scientific articles proposing mobile computer vision-based food analysis applications.
• Analysis of application functionalities focusing on food recognition, volume estimation, and calorific calculation.
• Evaluation of the presence and quality of explanations provided for classifications and estimations.

Main Results:

• A significant majority (90.9%) of reviewed applications failed to differentiate between food and non-food items.
• Only one study incorporated explanations for the features influencing classification in a dietary intake application.
• Lack of explainability hinders user trust and understanding of the system's outputs.

Conclusions:

• Mobile computer vision applications show promise for dietary monitoring and chronic illness management (e.g., diabetes).
• Current applications often lack basic food identification capabilities and transparency in their estimations.
• To enhance trust and utility in healthcare, these applications must provide clear explanations for their classifications and estimations.