Personalized Wedding Photo Generator (Server)

Backend service for the NUS MSBA Capstone project: Personalized Wedding Photo Generation Using Generative AI.
Built with Flask (Python), integrating Stable Diffusion, IP-Adapter FaceID, InsightFace, and BLIP.

---

Getting Started

# Create virtual environment (optional)
python -m venv venv
source venv/bin/activate   # macOS/Linux
venv\Scripts\activate      # Windows

# Install dependencies
pip install -r requirements.txt

# Run development server
python server.py

Server runs at http://localhost:5000.

---

Project Structure

app/                # Application layer
│   ├── photo_generator.py  # Photo generation workflow
│   └── server.py           # Flask service entry
ip_adapter/         # IP-Adapter core modules
models/             # Model weights & configs
public/             # Generated outputs
uploaded/           # User-uploaded files

---

API Endpoints

• POST /upload
  Upload face and background images.

• POST /prepare
  Extract identity embeddings, analyse background.

• POST /generate
  Generate personalized wedding photo.

---

Features

• Face Embedding Extraction: Powered by InsightFace\
• Background Harmonisation: BLIP caption + INR-Harmonization\
• Identity Preservation: IP-Adapter FaceID\
• Generation Pipeline: Stable Diffusion (Realistic Vision) + VAE\
• Modular Design: Preprocessing, prompt construction, and generation are independent

---

Screenshots

Architecture diagram: