🎬 Celebrity Look-Alike Finder

Find out which Bollywood celebrity or sports personality you look like using advanced facial recognition technology!

🌟 Features

• Face Detection: Uses MTCNN for accurate face detection
• Deep Learning: VGGFace with ResNet50 for feature extraction
• High Accuracy: Compares with multiple celebrity images
• Top 5 Matches: Shows your top 5 celebrity matches
• Confidence Scores: Displays match confidence percentages
• Wide Celebrity Database: Includes Bollywood actors and Indian sportspersons

📋 Prerequisites

• Python 3.10 (Required for TensorFlow compatibility)
• Webcam or image files for testing

🚀 Quick Start (Local)

1. Install Dependencies

pip install -r requirements.txt

2. Run the Application

streamlit run celebrity_finder.py

The app will open in your browser at http://localhost:8501

☁️ Deploy on Render

Step-by-Step Deployment Guide

1. Prepare Your Repository

# Initialize git (if not already done)
git init
git add .
git commit -m "Initial commit for Render deployment"

2. Create a GitHub Repository

1. Go to github.com/new
2. Create a new repository (e.g., celebrity-lookalike-finder)
3. Push your code:

git remote add origin https://github.com/YOUR_USERNAME/celebrity-lookalike-finder.git
git branch -M main
git push -u origin main

3. Set Up Render Account

1. Go to render.com
2. Sign up with GitHub (recommended for easy integration)
3. Click "Dashboard"

4. Create a New Web Service

1. Click "New +" → "Web Service"
2. Select "Build and deploy from a Git repository"
3. Click "Connect" next to your GitHub repository
4. Select the repository: celebrity-lookalike-finder

5. Configure the Service

Basic Settings:

• Name: celebrity-lookalike-finder
• Environment: Python 3
• Build Command: pip install --no-cache-dir -r requirements.txt
• Start Command: streamlit run celebrity_finder.py --server.port=8501 --server.address=0.0.0.0

Instance Type: Free (for testing)

6. Set Environment Variables

Add these environment variables in Render dashboard:

PYTHONUNBUFFERED=1

7. Deploy

1. Click "Create Web Service"
2. Wait for deployment (2-5 minutes)
3. Your app will be live at: https://celebrity-lookalike-finder.onrender.com

Alternative: Using render.yaml

The render.yaml file is already configured. You can also deploy by:

1. Going to render.com/dashboard
2. Click "New +" → "Web Service"
3. Select "Build and deploy from Git"
4. Render will automatically use render.yaml configuration

📁 Project Structure

├── celebrity_finder.py          # Main Streamlit application
├── app.py                       # Alternative app (older version)
├── extract_features.py          # Feature extraction pipeline
├── download_celebrity_images.py # Image downloader setup
├── requirements.txt             # Python dependencies
├── Procfile                     # Render deployment config
├── render.yaml                  # Render service config
├── .streamlit/config.toml       # Streamlit configuration
├── README.md                    # This file
├── celebrity_images/            # Celebrity image database
├── uploads/                     # User uploaded images
├── embedding.pkl                # Celebrity facial embeddings
└── filenames.pkl               # Celebrity image filenames

🎯 How It Works

1. Face Detection: MTCNN detects faces in the uploaded image
2. Preprocessing: Face is extracted, padded, and resized to 224x224
3. Feature Extraction: VGGFace ResNet50 extracts 2048-dimensional features
4. Similarity Matching: Cosine similarity compares with all celebrity embeddings
5. Ranking: Top 5 most similar celebrities are displayed with confidence scores

🔧 Improving Accuracy

To improve the model's accuracy:

1. More Images: Add 30-50 images per celebrity
2. Diverse Angles: Include various facial angles and expressions
3. Good Quality: Use high-resolution, clear images
4. Proper Lighting: Ensure images have good lighting
5. Face Clarity: Make sure faces are clearly visible

📊 Celebrity Database

Current database includes Bollywood actors and Indian sports personalities.

🛠️ Troubleshooting

Deployment Issues

"Dependencies failed to install"

• Ensure Python 3.10 is specified
• Check that all packages in requirements.txt are compatible
• Try removing large packages temporarily

"App crashes on startup"

• Check logs in Render dashboard
• Ensure embedding.pkl and filenames.pkl files are present
• Verify TensorFlow/Keras versions match

"No face detected"

• Ensure your face is clearly visible
• Use good lighting
• Face the camera directly
• Avoid sunglasses or face coverings

Memory Issues

If you get memory errors:

• Use Render's paid plans with more RAM
• Optimize image processing in extract_features.py
• Reduce number of celebrity images processed at once

📝 Environment Variables

Set these in Render dashboard if needed:

PYTHONUNBUFFERED=1          # Shows print output immediately
STREAMLIT_SERVER_PORT=8501  # Port for Streamlit

🤝 Contributing

To add more celebrities:

1. Create folder: celebrity_images/category/Celebrity_Name/
2. Add 20-50 clear face images
3. Run: python extract_features.py (locally)
4. Commit and push changes
5. Render will auto-redeploy

📧 Support

For issues and questions:

1. Check the README troubleshooting section
2. Review error messages in the Render logs
3. Ensure all files (embedding.pkl, filenames.pkl) are included

🔗 Useful Links

• Streamlit Documentation
• Render Documentation
• TensorFlow/Keras Documentation
• MTCNN GitHub

---

Note: This is a fun project for educational purposes. Results may vary based on the quality and quantity of celebrity images in your database.