FaceSearch

A deep learning-based face search application using two different models: MobileFaceNet and InsightFace.

Features

• Download images from Google Drive using PyDrive.
• Detect and extract faces using MTCNN or InsightFace's built-in detector.
• Generate face embeddings with either MobileFaceNet or InsightFace.
• Search for similar faces in a directory using cosine similarity.
• Save matched results with bounding boxes.

Model Usage

• MobileFaceNet:
  Used in searchUsingMobileNet.py. This script uses MTCNN for face detection and MobileFaceNet for generating face embeddings. It is suitable for lightweight, efficient face recognition tasks.

• InsightFace:
  Used in searchUsingInsightFace.py. This script leverages the InsightFace library for both face detection and embedding extraction, providing robust and accurate face recognition.

Requirements

• Python 3.8+
• PyTorch
• torchvision
• scikit-learn
• opencv-python
• mtcnn
• insightface
• PyDrive

Setup

1. Clone the repository:

   git clone https://github.com/your-username/FaceSearch.git
   cd FaceSearch

2. Install dependencies:

   pip install -r requirements.txt

3. Google Drive API setup:

   • Place your client_secrets.json in the project directory.
   • Enable the Google Drive API for your Google Cloud project.

4. Download images from Google Drive:

   python downloadPydrive.py

5. Run face search with MobileFaceNet:

   python searchUsingMobileNet.py

6. Run face search with InsightFace:

   python faceSearch.py

File Structure

• downloadPydrive.py - Downloads images from a Google Drive folder.
• searchUsingMobileNet.py - Searches for similar faces using MobileFaceNet.
• searchUsingInsightFace.py - Searches for similar faces using InsightFace.
• mobileFaceNet.py - MobileFaceNet model definition.
• client_secrets.json - Google API credentials (not tracked by git).
• output/ - Saved images with matched faces.

Notes

• Make sure to update reference_image_path and search_directory in the scripts as needed.
• Model checkpoint file (*.ckpt) should be present in the project directory for MobileFaceNet.

License

This project is for educational purposes.