rakshittoast/food-aesthetic

🍽️ Food Aesthetic Scoring Service

A machine learning service that automatically scores food images based on their aesthetic quality using a pre-trained MobileNet model.

🌟 Features

• —AI-Powered Scoring: Uses a fine-tuned MobileNet model to evaluate food aesthetics
• —RESTful API: FastAPI-based service with multiple endpoints
• —Batch Processing: Score multiple images at once
• —Multiple Input Formats: Support for file uploads and base64 encoded images
• —Real-time Scoring: Get aesthetic scores in milliseconds
• —CORS Enabled: Ready for web applications

🚀 Quick Start

Prerequisites

• —Python 3.8+
• —TensorFlow 2.9
• —4GB+ RAM recommended

Installation

1. 1.Clone the repository

   git clone <your-repo-url>
   cd food-aesthetic-scoring-service

1. 1.Install dependencies

   pip install -r requirements.txt

1. 1.Start the API server

   ./start_api.sh

Or manually:

   python api.py

1. 1.Test the service

   python demo.py

📡 API Endpoints

Health Check

GET /health

Check if the model is loaded and ready.

Score Single Image

POST /score

Score a single food image. Accepts both file uploads and base64 encoded data.

File Upload:

curl -X POST "http://localhost:8000/score" \
     -H "accept: application/json" \
     -H "Content-Type: multipart/form-data" \
     -F "file=@your_food_image.jpg"

Base64 Data:

curl -X POST "http://localhost:8000/score" \
     -H "Content-Type: application/json" \
     -d '{"image_data": "base64_encoded_string", "image_format": "jpeg"}'

Batch Scoring

POST /score-batch

Score multiple images at once for efficient processing.

API Information

GET /

Get API version and available endpoints.

🧠 Model Details

The service uses a MobileNet-based neural network that has been specifically trained on food images to predict aesthetic quality. The model:

• —Takes 224x224 RGB images as input
• —Outputs scores between 0 and 1 (higher = more aesthetic)
• —Uses temperature scaling for calibrated predictions
• —Processes images with aspect ratio preservation

Score Interpretation

💻 Usage Examples

Python Client

import requests
from PIL import Image
import base64
import io

# Initialize client
base_url = "http://localhost:8000"

# Score an image file
with open("food_image.jpg", "rb") as f:
    files = {"file": ("image.jpg", f, "image/jpeg")}
    response = requests.post(f"{base_url}/score", files=files)
    
if response.status_code == 200:
    score = response.json()["aesthetic_score"]
    print(f"Aesthetic Score: {score:.4f}")

Base64 Encoding

# Convert image to base64
with open("food_image.jpg", "rb") as f:
    image_bytes = f.read()
    base64_data = base64.b64encode(image_bytes).decode()

# Send to API
data = {
    "image_data": base64_data,
    "image_format": "jpeg"
}
response = requests.post(f"{base_url}/score", json=data)

🛠️ Development

Project Structure

food-aesthetic-scoring-service/
├── api.py                 # FastAPI server
├── food_aesthetics/       # Core ML model
│   ├── model.py          # FoodAesthetics class
│   └── trained_weights.h5 # Pre-trained model weights
├── demo.py              # Usage examples
├── requirements.txt      # Python dependencies
└── start_api.sh         # Server startup script

Running Tests

# Test the API endpoints
python test_api.py

# Test base64 functionality
python test_base64_api.py

Model Training

The model is pre-trained and ready to use. If you need to retrain:

1. 1.Prepare your food image dataset
2. 2.Modify the FoodAesthetics class in model.py
3. 3.Train using TensorFlow/Keras
4. 4.Save weights to trained_weights.h5

🔧 Configuration

Environment Variables

• —PORT: API server port (default: 8000)
• —HOST: API server host (default: 0.0.0.0)

Model Parameters

• —Input Size: 224x224 pixels
• —Batch Size: 1 (configurable in model.py)
• —Temperature: 1.537 (for score calibration)

📊 Performance

• —Inference Time: ~50-100ms per image
• —Memory Usage: ~2-3GB RAM
• —Concurrent Requests: Supports multiple simultaneous users
• —Model Loading: ~5-10 seconds on first startup

🤝 Contributing

1. 1.Fork the repository
2. 2.Create a feature branch
3. 3.Make your changes
4. 4.Add tests if applicable
5. 5.Submit a pull request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• —Built with TensorFlow and FastAPI
• —Model architecture based on MobileNet
• —Food aesthetics dataset for training

📞 Support

For questions or issues:

• —Open an issue on GitHub
• —Check the API documentation
• —Review the demo scripts for usage examples

Happy Food Aesthetic Scoring! 🍕🍜🍰