The challenge

Manual food logging in traditional fitness apps is tedious and prone to human error, causing over 70% of users to abandon dietary tracking within their first week.

The approach

Multimodal Visual Recognition: Integrated GPT-4 Vision with custom few-shot dietary prompting to detect complex Asian and Western cuisine dishes. Portion Size Estimation: Used reference object detection to calculate accurate portion sizing and calorie density. Personalized Macro Targets: Automatically adjusted daily calorie and nutrient goals based on user metabolic rate and fitness milestones.

What we did

• //001Snap & Log: Complete meal analysis generated in under 2.5 seconds
• //002Barcode Scanner Fallback
• //003Weekly Progress Analytics