Dlib

Dlib is a modern C++ toolkit of machine learning algorithms and supporting tools for building complex software that solves real world problems. It is a mature, general purpose library rather than a hosted service: developers compile it into their own applications and get classical machine learning, deep learning, computer vision and numerical tools in one dependency, with Python bindings for much of the functionality. It is used in industry and academia across robotics, embedded devices, mobile phones and high performance computing. The machine learning side covers support vector machines for classification, regression and ranking, relevance vector machines, kernel methods, clustering, structural prediction, sequence labelling and a deep learning toolkit for image classification, object detection, semantic and instance segmentation, metric learning and language modelling. Its computer vision tools are widely used for face detection, facial landmark detection, face recognition, object tracking and training YOLO style detectors. Beyond machine learning it includes linear algebra, optimisation, Bayesian networks, graph tools, image processing, compression, networking, threading and FFmpeg based video capture and encoding. The project stands out for complete documentation of every class and function, plus debugging modes that check documented preconditions. Dlib is free of charge for any application. It is released under the Boost Software License, which permits commercial use, and the current release listed on the site is version 20.0. There is no paid edition or support plan. Development takes place on GitHub, and the site carries extensive C++ and Python example programs for each area of the library.