Face Cluster

This project uses Docker to manage dependencies and containerize the environment, enabling easy deployment and execution of facial recognition and clustering tasks.

Prerequisites:

Ensure you have the following installed:

• Docker installed on your system.
• Docker Compose for managing multi-container applications.

Libs:

The system leverages Python libraries such as:

• OpenCV: Utilized for image processing and face detection.
• Scikit-learn: Used for clustering the facial encodings using algorithms like DBSCAN.
• Imutils: A utility library to assist with common image-processing tasks.
• Numpy: Essential for numerical operations, especially when handling large sets of encodings.
• Face Recognition: A deep learning-based library for detecting and encoding faces.

Installation:

Follow these steps to get the project up and running:

1. Clone repository

Use this to copy the source code in your machine:

git clone https://github.com/felipeversiane/face-cluster.git

2. Build and running:

Use this to build, download the dependences and running app :

docker-compose up --build

With this command, the code will be executed, first running the encode_faces.py file and then the cluster_faces.py file. This will create two folders called results and encodings, in which you will have the result of the executed code.

Directory structure:

• /src : Contains the Python source code for face encoding encode_faces.py and clustering cluster_faces.py.

• /database : Folder for storing the input images used for generating face encodings and clusters.

• encodings/ : Directory where the serialized facial encodings will be saved after processing.

• results/ : The output folder where clustered images and results will be saved.

Results:

The cluster results will be stored in the results folder, where it will be possible to see each label/cluster created with their respective images

Example:

• label1: image0.jpg, image1.jpg, image14.jpg ....
• label2: ....

Environment Variables:

• ENCODINGS_PATH: Path to the serialized encodings file.

• FACE_DATA_PATH: Path to store face data during processing.

• CLUSTERING_RESULT_PATH: Path where clustering results will be saved.

Usage in Docker:

• encode-faces processes images and generates face encodings.

• cluster-faces uses DBSCAN to group faces into clusters, storing them in the designated result path.

Notes:

• The project is designed to be extensible, allowing additional services like a web interface for image upload or real-time face recognition.

• The current focus is on the core functionality of encoding and clustering faces.