🍔 Food Image Classification using PyTorch + EfficientNet + Gradio

A deep learning project that classifies food images into 7 categories using a CNN-based EfficientNet model built with PyTorch.

The system provides a simple Gradio web interface where users can upload an image and instantly get predictions.

---

🚀 Features

✅ Image classification using EfficientNet
✅ 7 food categories
✅ Upload image and predict instantly
✅ Clean Gradio web interface
✅ GPU supported training (CUDA)
✅ Lightweight and fast inference

---

🧠 Classes Detected

The model classifies images into:

• 🥟 Samosa
• 🍕 Pizza
• 🥩 Steak
• 🍣 Sushi
• 🧁 Cup Cakes
• 🍟 French Fries
• 🍳 Omelette

---

🛠 Tech Stack

Tool	Purpose
PyTorch	Model training & inference
EfficientNet	CNN architecture
Torchvision	Pretrained models & transforms
Gradio	Web UI for predictions
CUDA	GPU acceleration

---

📂 Project Structure

food-classifier/
│
├── dataset/              # food images dataset
├── models/               # saved trained weights (.pth)
├── train.py              # training script
├── predict.py            # inference code
├── app.py                # gradio interface
├── requirements.txt
└── README.md

---

⚙️ Installation

1️⃣ Clone repository

git clone https://github.com/yourusername/food-classifier.git
cd food-classifier

---

2️⃣ Create virtual environment (recommended)

python -m venv venv
source venv/bin/activate      # Linux/Mac
venv\Scripts\activate         # Windows

---

3️⃣ Install dependencies

pip install -r requirements.txt

or manually:

pip install torch torchvision gradio pillow opencv-python numpy

---

🏋️ Training the Model

python train.py

Training Details

• Architecture: EfficientNet (CNN)
• Framework: PyTorch
• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Classes: 7
• GPU supported (CUDA)

The trained model will be saved inside:

models/model.pth

---

▶️ Run the Web App (Gradio)

python app.py

Then open in browser:

http://127.0.0.1:7860

---

🖼 How It Works

1. Upload an image
2. Image is preprocessed
3. EfficientNet extracts features
4. Model predicts class
5. Label + confidence shown on screen

---

💡 Example Usage

Upload:

📸 samosa.jpg

Output:

Prediction: Samosa
Confidence: 96.3%

---

📊 Model Architecture

EfficientNet is used because:

✅ High accuracy
✅ Fewer parameters
✅ Faster training
✅ Good for small datasets

---

🔮 Future Improvements

• Add webcam support
• Add top-3 predictions
• Mobile deployment
• Convert to TorchScript
• Deploy on HuggingFace Spaces
• Add more food classes

---

🤝 Contributing

Pull requests are welcome.
Feel free to open issues for improvements or suggestions.

---

📜 License

MIT License

---

👤 Author

Anshu Singh
ML Enthusiast

Built as a hands-on CNN + PyTorch learning project 🚀