Description
You want face matching in your own app, but cloud APIs bill per call and the photos have to go to someone else's server. That is where most attendance, door access and photo library projects get stuck.
Open Source Face Recognition SDK is Faceplugin's open-source toolkit, and all processing happens on your own machine, so the data never leaves the device. A few lines of Python find faces in an image, extract feature vectors and compare two faces, with no license fee and no GPU required; an ordinary CPU is enough.
It supports Windows and Linux and needs Python 3.9 or newer. To check whether two photos show the same person, the example uses a similarity threshold of 75, which you can use as is or tune for your own case.
Face detection: finds faces in an image and returns bounding boxes, with a cap on how many faces to detect.
Landmarks: extracts facial landmarks for alignment and further processing.
Feature embeddings: generates an embedding for each face, so you can store it and compare repeatedly without recomputing.
Similarity comparison: returns a similarity percentage for two faces to decide whether they are the same person.
Fully local: everything runs locally, so no private data is uploaded, and it works on a CPU.
Common image formats: reads JPG, PNG and other usual formats directly.
Open Source Face Recognition SDK is Faceplugin's open-source toolkit, and all processing happens on your own machine, so the data never leaves the device. A few lines of Python find faces in an image, extract feature vectors and compare two faces, with no license fee and no GPU required; an ordinary CPU is enough.
It supports Windows and Linux and needs Python 3.9 or newer. To check whether two photos show the same person, the example uses a similarity threshold of 75, which you can use as is or tune for your own case.
Features
Face detection: finds faces in an image and returns bounding boxes, with a cap on how many faces to detect.
Landmarks: extracts facial landmarks for alignment and further processing.
Feature embeddings: generates an embedding for each face, so you can store it and compare repeatedly without recomputing.
Similarity comparison: returns a similarity percentage for two faces to decide whether they are the same person.
Fully local: everything runs locally, so no private data is uploaded, and it works on a CPU.
Common image formats: reads JPG, PNG and other usual formats directly.
