Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

nataphong-lab/Open-Source-Face-Recognition-Ultimate

Machine Learning

A local Python SDK for face detection, landmark and embedding extraction, and face similarity comparison on Windows and Linux.

459
369
Python
1 day old

What it is

The project packages a face recognition pipeline into a Python interface. It detects one or more faces in an image, returns bounding boxes and landmarks, extracts feature embeddings, and compares two embeddings to produce a 0–100 similarity score. All processing is described as running locally on Windows or Linux, on CPU or with optional GPU acceleration.

The recognition model is based on FacePlugin's open-source model and was further trained by this repository's author. The README reports benchmark comparisons against FacePlugin on LFW, CFP-FP, AgeDB-30, CALFW, and CPLFW, with an average of 98.28 versus 83.06; these are the author's own evaluation figures, not independent results README.

Quick start

The documented workflow targets Windows:

  1. Clone the repository and enter its directory:

    git clone https://github.com/nataphong-lab/Open-Source-Face-Recognition-Ultimate.git
    cd Open-Source-Face-Recognition-Ultimate
    
  2. Install Python 3.9 and run the repository's setup script from the repository directory:

    winget install --id Python.Python.3.9 -e --source winget
    powershell -Ep Bypass ./install.ps1
    
  3. Create and activate a Conda environment, then install Python dependencies:

    conda create -n facesdk python=3.9 -y
    conda activate facesdk
    pip install -r requirements.txt
    python run.py
    

For library use, import FaceRecognition, call GetImageInfo(image_path, faceMaxCount=10), and read bbox, landmarks, and embedding from each returned face. To compare two faces, extract one embedding from each image and pass them to get_similarity; the README uses a similarity of 75 as an example same-person threshold and notes it must be validated for your own data and conditions usage example.

The README does not explain what install.ps1 does or how it prepares the environment, so that step cannot be reproduced from the provided material alone installation.

Constraints and practical notes

  • Python 3.9 or later is required; Anaconda or Miniconda is recommended, and the documented setup uses Conda.
  • Supported image formats are JPG, PNG, BMP, and TIFF. The API accepts a file path and a maximum face count.
  • The repository declares no license, so redistribution and commercial use terms are unclear.
  • The reported benchmark numbers come from the project's own comparison and do not show the test protocol, preprocessing, or evaluation code in the README.
  • run.py is listed as the entry point, but the README does not describe the resulting interface or output.

HotRepo's take

The project entered HotRepo's new-repository list on 2026-11-19, recording 110 stars in roughly 13 hours and 82 forks in the same window. That is a fast opening for a repository whose documented surface is a Python face-recognition API and whose main setup path includes a PowerShell script. The provided HotRepo data has no daily all-language rank for the repository, so the record supports visibility on that day, not a position on the GitHub-wide daily chart or a sustained trend.

Written by AI from the project's README, updated . The project's own docs are the source of truth.

You might also like

  1. inclusionAI/IDMResearch code and inference toolkit for one- or two-step text-to-image generation via intrinsic distribution matching.Python161
  2. speridlabs/idol-sdkTurns a single object image into a standard 3D Gaussian splat on consumer GPUs, for 3D artists, researchers and developers.Python385
  3. ubc-vision/solv3rA pretrained multi-view 3D reconstruction model that solves camera poses and point maps from image or video sequences.Python158
  4. pookjw/HermesEnablerEnables Hermès watch faces on Apple Watch by sideloading an app and modifying device configuration.Objective-C1,316
  5. isl-org/Open3DA C++ and Python library for working with 3D data: point clouds, meshes, registration, reconstruction and visualization.C++14,124