OpenCV

Vision software across the AI workflow

OpenCV provides computer-vision and machine-learning software for processing images and video, extracting useful features and running supported neural networks. That spans conventional vision algorithms and learned models. The project is operated by the Open Source Vision Foundation, whose organizational identity is distinct from the software community recorded here. [1]

OpenCV 4.0 shipped in November 2018. The release moved to C++11, expanded the deep-neural-network module and ONNX model support, and introduced G-API graph construction. Its experimental Vulkan backend illustrates a deployment direction, rather than proof that every neural network ran on every GPU. [2][3]

The licensing history also matters when using older versions. OpenCV 4.5.0 and later use Apache-2.0, while version 4.4.0 and earlier, including the original 4.0 release, use BSD 3-clause terms. Model weights and third-party components can introduce additional conditions beyond the main library's license. [4]