Food Vision 101

A PyTorch-based image classification project using EfficientNetB2 and Vision Transformer (ViT) to classify food images into pizza, steak, or sushi. The final model is deployed using Gradio and hosted on Hugging Face Spaces.

---

🚀 Project Structure

FOODVISION-101/
│
├── data/                           # Raw and extracted datasets
│   ├── pizza_steak_sushi/          # Preprocessed food image dataset
|   └── food-101.zip                # Original ZIP archive
│
├── models/                         # Trained PyTorch model weights (.pth)
│   ├── effnetb2.pth
│   ├── effnetb2_food101.pth
│   ├── tinyvgg.pth
│   └── vit.pth
│
├── examples/                       # Example images used by Gradio app
│
├── notebooks/                      # Jupyter notebooks for EDA, training, testing
│   ├── 1.0-rs-exploratory-data-analysis.ipynb
│   ├── 2.0-rs-train-eval-models.ipynb
│   └── 3.0-rs-interface-test.ipynb
│   └── 4.0-rs-foodvision-101.ipynb
│
├── reports/                        # Visualizations and evaluation reports
│   └── figures/
│       └── foodvision-mini-inference-speed-vs-performance.jpg
│
├── src/                            # Source code (modular structure)
│   ├── config.py                    # Global config variables (paths, class names)
│   ├── dataset.py                  # Data download/unzipping helpers
│   ├── features.py                 # Transforms and DataLoader creation
│   ├── plots.py                    # Visualization utilities
│   │
│   ├── modeling/                   # Model logic
│   │   ├── models.py               # Model creation: EffNetB2, ViT, etc.
│   │   ├── train.py                # Training loop
│   │   ├── predict.py              # Inference helpers
│   │   └── utils.py                # Save/load model, metrics, etc.
│   │
│   ├── services/                   # External integrations (e.g. Gradio app)
│   │   └── demo.py                 # Gradio Blocks interface
│   │
│   └── scripts/                    # CLI scripts to automate training/testing
│       ├── generate_examples.py    # Generate and copy example images
│       ├── train_effnetb2.py       # Train EfficientNetB2 model
│       └── train_vit.py            # Train ViT model
│
├── .gitignore
├── app.py                          # Launch Gradio app (entry point)
├── requirements.txt                # Python package dependencies
└── README.md

---

🧠 Key Features

• 🧠 Transfer learning with EfficientNetB2 and ViT
• 🧹 Clean modular code (src/ structure)
• 📈 Training via script (src/scripts/train_effnetb2.py, train_vit.py)
• 🖼️ Real-time image predictions with Gradio
• ☁️ Deployment-ready for Hugging Face Spaces

---

📦 Installation

git clone git@github.com:RahulSaini02/pytorch-foodvision-101.git
cd foodvision-101
python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows
pip install -r requirements.txt

Run Locally

python src/scripts/generative_examples.py

python src/scripts/train_effnetb2.py
# or
python src/scripts/train_vit.py

Launch Gradio App

python app.py