ageitgey/face_recognition

Face Recognition is a Python library that makes it straightforward to find, identify, and compare human faces in images and video. The goal is to give developers the simplest possible interface for facial recognition tasks without needing to understand the underlying deep learning machinery. The library solves a practical problem: facial recognition is technically complex, normally requiring knowledge of convolutional neural networks, computer vision pipelines, and model training. This library wraps all of that complexity so you can detect, identify, and compare faces in just a few lines of code. How it works under the hood: the library is built on top of dlib, a C++ machine learning library with pre-trained deep learning models. When you pass an image to the library, it first locates faces in the image by finding the characteristic patterns of eyes, nose, and mouth. Then it generates a numerical "encoding", a fixed-length list of 128 numbers, that represents each face in a way that is stable across different photos of the same person. To recognize a person, you compare their encoding against a set of known encodings, if the distance between them is small enough, it's a match. This approach achieves 99.38% accuracy on the standard Labeled Faces in the Wild benchmark. You can also extract facial landmark points, the precise locations of eyes, eyebrows, nose, lips, and chin, for more detailed work like applying digital filters. You would use this library for building attendance systems, security camera analysis, photo organization by person, or any application that needs to answer "who is in this image?" The tech stack is Python, with dlib and cmake as dependencies, and it also includes a command-line tool for batch processing folders of images without writing any code.