🏛️ Architectural Style Classifier

A deep learning model that classifies buildings into 25 architectural styles. Built with PyTorch and EfficientNet, deployed as both an interactive Gradio demo and a REST API.

🎯 Live Demo

Try it now: Hugging Face Space

Upload any building image and get instant predictions across 25 architectural styles.

📸 Example

Input	Prediction
	Gothic (100%)
	Tudor Revival (95%)
	Art Deco (98%)

🏗️ Architectural Styles

The model recognizes 25 styles:

Achaemenid	American Craftsman	American Foursquare	Ancient Egyptian	Art Deco
Art Nouveau	Baroque	Bauhaus	Beaux-Arts	Byzantine
Chicago School	Colonial	Deconstructivism	Edwardian	Georgian
Gothic	Greek Revival	International	Novelty	Palladian
Postmodern	Queen Anne	Romanesque	Russian Revival	Tudor

🚀 Quick Start

Option 1: Use the API

import requests

response = requests.post(
    "https://architectural-style-classifier.onrender.com/predict",
    files={"file": (os.path.basename(image_path), open("building.jpg", "rb"), "image/jpg")}
)

result = response.json()
print(result["top_style"])  # "Gothic"
print(result["predictions"])  # {"Gothic": 0.892, "Romanesque": 0.056, ...}

Option 2: Run Locally

# Clone the repo
git clone https://github.com/ZoeCD/architectural-style-classifier.git
cd architectural-style-classifier

# Install dependencies
pip install -r requirements.txt

# Run the Gradio app
python app/gradio_app.py

Then open http://localhost:7860 in your browser.

Option 3: Docker

docker build -t arch-classifier .
docker run -p 8000:8000 arch-classifier

API available at http://localhost:8000. Interactive docs at http://localhost:8000/docs.

📁 Project Structure

architectural-style-classifier/
├── app/
│   ├── gradio_app.py      # Hugging Face Spaces demo
│   └── main.py            # FastAPI REST API
├── src/
│   ├── train.py           # Training script
│   └── predict.py         # Inference module
├── models/
│   ├── arch_classifier_best.pth
│   └── class_labels.json
├── notebooks/
│   └── data-exploration.ipynb
|   └── example-api.ipynb  # API request example
├── example_images/              # Sample images for demo
├── Dockerfile
├── requirements.txt
└── README.md

🔧 API Reference

GET /

Health check endpoint.

Response:

{"message": "Architectural Style Classifier API", "status": "running"}

GET /health

Service health status.

Response:

{"status": "healthy"}

POST /predict

Classify an architectural image.

Request: Multipart form with image file

Response:

{
  "predictions": {
    "Gothic": 0.892,
    "Romanesque": 0.056,
    "Byzantine": 0.023,
    "Baroque": 0.012,
    "Art Nouveau": 0.008
  },
  "top_style": "Gothic"
}

🧠 Model Details

Attribute	Value
Architecture	EfficientNet-B0
Pretrained on	ImageNet
Fine-tuned on	15,432 architectural images
Input size	224 × 224
Output	25 classes
Test accuracy	~70%

Training Approach

1. Transfer learning: Started with ImageNet-pretrained EfficientNet-B0
2. Two-phase training:
   • Phase 1: Trained classifier head only
   • Phase 2: Fine-tuned entire network with lower learning rate
3. Class imbalance handling: Weighted cross-entropy loss
4. Data augmentation: Random flips, rotation, color jitter

📊 Dataset

Based on the Architectural Styles Dataset from Kaggle, which combines:

• Original dataset from "Architectural Style Classification using Multinomial Latent Logistic Regression" (ECCV 2014) by Zhe Xu et al.
• Additional images scraped from Google Images

🙏 Acknowledgments

• Dataset by dumitrux and Zhe Xu et al.
• EfficientNet implementation from torchvision
• Deployed with Hugging Face Spaces and Render