CompreFace: Run Self-Hosted Face Recognition APIs

Overview

CompreFace packages face analysis as a self-hosted service rather than requiring teams to build and operate machine-learning models themselves. Applications can use its REST API for face recognition, verification, and detection, alongside additional face-analysis functions.

It is aimed at developers and teams who want a deployable service for on-premises or cloud environments, with Docker Compose as the documented starting point. CPU and GPU model options are supported, and a web interface provides service and role management.

Key Features

• REST APIs for face recognition, verification, and detection.
• Additional functions include landmark detection, mask detection, head pose, age, and gender recognition.
• Docker Compose deployment for running the service as a standalone server.
• CPU and GPU model support.
• Web interface with user roles and access management.
• Uses FaceNet and InsightFace libraries.
• SDKs are available for JavaScript, Python, and .NET.

Use Cases

• Application developers can add face verification or identification through an API instead of building a recognition service from scratch.
• Organizations that need face processing on premises can run the service on infrastructure they control.
• Teams prototyping computer-vision workflows can use the API and UI without prior machine-learning expertise.
• Integrators can connect face-analysis capabilities to existing systems using REST APIs or the listed language SDKs.

Getting Started

Install Docker and Docker Compose, download and unzip an archive from the releases page, then run:

docker-compose up -d

Open http://localhost:8000/login to access the service. The README and documentation cover setup and configuration details.

Considerations

• Docker and Docker Compose are required for the documented deployment path.
• The README specifies an x86 processor with AVX support as the expected hardware baseline. GPU support is available, but is not stated as a requirement.
• Face recognition results and suitability depend on the selected model and application requirements. Review the project's API and configuration documentation before deployment.
• The repository is not archived, but its latest push in the provided metadata is dated 2024-10-05, so check the repository and releases for current maintenance and compatibility information.