Skip to content

Latest commit

 

History

History
243 lines (169 loc) · 6.58 KB

README.md

File metadata and controls

243 lines (169 loc) · 6.58 KB

🥗 Food Calorie Estimator AI

🍔 Deep Learning App to Estimate Food Calories Trained With 101,000+ Images Dataset

![Food Calorie Estimator]

🚀 AI-powered calorie prediction from food images — built using PyTorch, Gradio, and EfficientNet. Predicts food type and estimated calories in real-time, hosted on Render with a smooth, interactive UI.

🌟 Overview

This project is an end-to-end AI system that:

  1. Detects and classifies the food in an image 🍱
  2. Estimates its calorie content 🔥
  3. Displays the result through a beautiful web interface 🌐

The model was trained using the Food-101 dataset (over 101,000 images across 101 classes) and fine-tuned with custom USDA nutrition data for calorie estimation.

🧠 Core Features

✅ Food classification using a fine-tuned EfficientNet-B0 CNN ✅ Calorie prediction using USDA-based mapping ✅ Clean and interactive Gradio interface ✅ Fully deployable on Render (or Hugging Face, Colab, etc.) ✅ Lightweight and optimized for CPU inference

🏗️ Project Architecture

📦 food-calorie-ai
├── app.py                        # Main Gradio application
├── requirements.txt              # Dependencies
├── data/
│   ├── class_calorie_mapping.json   # Food class → Calorie mapping
│   ├── metadata.csv                # Training metadata
│   ├── models/
│   │   ├── mh_net.pth              # PyTorch model (for retraining)
│   │   └── mh_net_traced.pt        # TorchScript model (for deployment)
│   └── usda/                       # USDA dataset folder
├── README.md
└── sample_images/                 # Example test images

🔬 Model Development Pipeline

Step 1. Dataset

  • Dataset Used: Food-101
  • 101 classes × 1000 images each
  • Cleaned and preprocessed using torchvision

Step 2. Preprocessing

transform = T.Compose([
    T.Resize((224,224)),
    T.ToTensor(),
    T.Normalize([0.485,0.456,0.406],[0.229,0.224,0.225])
])

Step 3. Model Architecture

  • Base: EfficientNet-B0 pretrained on ImageNet

  • Custom head for:

    • Softmax classification → Food Type
    • Linear regression → Calorie Prediction

Step 4. Training

optimizer = torch.optim.Adam(model.parameters(), lr=3e-4)
criterion_cls = torch.nn.CrossEntropyLoss()
criterion_cal = torch.nn.L1Loss()

✅ 8 Epochs on Colab Pro (with CUDA) ✅ Validation accuracy ~79.6% ✅ Mean Absolute Error ≈ 6.7 kcal

Step 5. Export

traced = torch.jit.trace(model, torch.randn(1,3,224,224))
torch.jit.save(traced, "data/models/mh_net_traced.pt")

🌐 Web App (Gradio + Render)

Frontend UI

![Gradio Interface Example]

Built using Gradio Interface:

demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil", label="Upload Food Image"),
    outputs=[gr.Markdown()],
    title="🍔 Food Calorie Estimator AI",
    description="Upload a food image to estimate its calorie content."
)

Deployment Command

python app.py

☁️ Render Deployment Guide

Step Command / Setting
1️⃣ Build Command pip install -r requirements.txt
2️⃣ Start Command python app.py
3️⃣ Port Render uses 8080 automatically
4️⃣ Environment Python 3
5️⃣ Public Link Appears once deployed ✅

Ensure your app.py ends with:

demo.launch(server_name="0.0.0.0", server_port=8080)

⚙️ Local Installation

1. Clone Repo

git clone https://github.com/YourUsername/food-calorie-ai.git
cd food-calorie-ai

2. Install Dependencies

pip install -r requirements.txt

3. Run App

python app.py

Open your browser at: 👉 http://127.0.0.1:7860

📊 Performance Metrics

Metric Value
Accuracy (Food Classification) 79.6%
MAE (Calorie Estimation) 6.76 kcal
Inference Speed (CPU) ~0.4 sec/image
Model Size 21 MB (EfficientNet-B0 traced)

📈 Future Enhancements

  • Add multi-food detection (object detection)
  • Integrate portion-size estimation using depth data
  • Train on custom Indian food dataset
  • Add voice assistant via Gradio chatbot
  • Deploy on mobile (Flutter/React Native) using ONNX model

🧩 Tech Stack

Category Tools
Language Python 3.10
Frameworks PyTorch, Torchvision
UI Gradio
Deployment Render
Dataset Food-101 + USDA Nutrition Data
Environment Google Colab (Training) + Render (Hosting)

📚 References

🪪 License

MIT License

Copyright (c) 2025 Shubham Mohite

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the “Software”), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

💡 About the Developer

👨‍💻 Shubham Mohite AI & Full-Stack Developer | Passionate about Computer Vision and Health-Tech 🌐 LinkedIn • GitHub

✨ If you like this project, please give it a ⭐ on GitHub! It helps others find it too.