Bite — AI-Powered Food Scanner (MVP)

Overview

Bite is an AI-powered food scanning application designed specifically for Singapore and Asian food products. Unlike existing Western-focused nutrition apps, Bite analyzes local packaged foods using ingredient-level understanding aligned with Health Promotion Board (HPB) dietary guidance.

The MVP validates the core pipeline:

• Barcode scanning
• Ingredient label OCR
• Ingredient understanding
• Health risk flagging
• Self-improving product database

This repository contains both the backend (AI + API) and a React Native frontend (Expo) for end-to-end demonstration.

---

Key Features (MVP)

1. Barcode Scanning

• Scan or manually enter food barcodes
• Retrieve product information from a Singapore-focused database
• Handle unknown products by creating pending entries

2. Ingredient Label Scanning (OCR)

• Capture ingredient labels via camera or photo upload
• OCR extracts raw ingredient text
• Designed for real-world packaging (varied fonts, lighting, layouts)

3. Ingredient Understanding & Health Analysis

• Normalizes ingredient text into structured components

• Flags potential health risks based on ingredient cues:

  • Added sugars
  • High sodium indicators
  • Saturated fat sources
  • Additives and allergens

• Produces non-medical, explainable health guidance

4. Self-Improving Database

• Products not found via barcode can be enriched using label scans
• Each scan improves coverage and accuracy over time
• Enables network effects as usage grows

---

System Architecture

Mobile App (React Native / Expo)
        ↓
Flask REST API (Backend)
        ↓
OCR (EasyOCR) + Ingredient Parsing
        ↓
Health Scoring Engine (HPB-aligned rules)
        ↓
SQLite Product & Ingredient Database

---

Repository Structure

Bite/
├── backend/
│   ├── app.py              # Flask API server
│   ├── db.py               # SQLite database logic
│   ├── scoring.py          # Ingredient normalization & health scoring
│   ├── ocr.py              # OCR pipeline
│   ├── seed.py             # Seed database script
│   ├── requirements.txt
│   └── data/
│       └── products_seed.csv
│
├── frontend/
│   ├── App.tsx             # React Native Expo app
│   ├── src/
│   │   ├── api.ts
│   │   └── config.ts
│   └── package.json
│
└── README.md

---

Backend Setup

Prerequisites

• Python 3.9+
• pip
• Virtual environment recommended

Installation

cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Initialize Database

python seed.py

This seeds the database with a small set of Singapore/Asian products for demo purposes.

Run Backend Server

python app.py

Backend runs at:

http://localhost:5000

---

Frontend Setup (React Native / Expo)

Prerequisites

• Node.js 18+
• Expo CLI
• Expo Go app (iOS / Android)

Installation

cd frontend
npm install

Configure Backend URL

Edit src/config.ts:

export const API_BASE = "http://<YOUR_LAPTOP_IP>:5000";

Note: localhost does not work on physical mobile devices. Use your computer’s local IP (same Wi-Fi network).

Run Frontend

npx expo start

Scan the QR code using Expo Go.

---

How to Use (Demo Flow)

Barcode Scan

1. Open the app
2. Scan a barcode using the camera
3. If product exists → view health analysis
4. If product is new → prompted to upload ingredient label

Ingredient Label Scan

1. Take or upload a photo of the ingredient list
2. OCR extracts ingredients
3. App displays health risk flags
4. Product is saved and enriched in the database

---

Health Scoring Disclaimer

Bite provides ingredient-based health guidance, not medical advice.

• No calorie counting or diagnosis
• No disease-specific recommendations
• Designed for awareness and informed choices

Users should always consult packaging labels and healthcare professionals when needed.

---

Technical Highlights

• OCR pipeline optimized for noisy, real-world images

• Rule-based ingredient interpretation (ML-ready)

• Clear upgrade path to:

  • YOLO-based ingredient panel detection
  • Multilingual OCR (Chinese/English)
  • Personalized dietary profiles

---

Future Work

• YOLO-powered ingredient region detection
• Expanded Singapore supermarket coverage
• FairPrice product data integration
• Premium features (personalized goals, history)
• Corporate wellness pilots