6. Food Detector

Overview

Snap or upload a photo of your plate and instantly identify foods and approximate portion sizes to estimate calories and macros.

Primary Use Cases

• Diet-tracking apps automating meal logs
• Nutritionists verifying client intake
• Individuals curious about “what am I eating?”

Key Features

• Image upload/camera widget in browser or mobile PWA
• Food recognition model returning multiple items (e.g. “scrambled eggs,” “bacon,” “toast”)
• Portion estimation via bounding-box area → rough volume
• Nutrition estimate using a lookup table
• History of meals with timestamps and totals

Tech Stack

• Frontend: React + TypeScript + Tailwind
• Backend: FastAPI (Python)
• AI Models:
  • Vision: google/vit-gcc fine-tuned on Food-101 or custom dataset
  • Portion sizing: simple regression model on known plate dimensions
• Storage: S3 or Firebase Storage for images; PostgreSQL for metadata

Architecture

1. Image Ingest → normalize size → send to vision model.