Gary Bradski, an Intel employee, comes up with an idea for a Computer Vision library. He creates a development team inside Intel and together they create what will become OpenCV. Vadim Pisarevsky becomes the technical lead of the library development team.

Based on Gary’s recommendation, Intel open-sources the code of the library, which was a huge milestone for Intel, the project and the team as the original plan was for a closed-source solution to be named CVL for “Computer Vision Library” . Gary came up with the name “OpenCV” that was inspired by the OpenGL framework that was already popular back then.

OpenCV was unveiled at CVPR’2000 conference held in Hilton Head Island, South Carolina, US. The library was highly acclaimed by the attendees of the conference and the community.

In 2005, OpenCV dev team was a part of the team that won the DARPA Grand Challenge in unmanned ground vehicle navigation. It required robots to navigate a 142-mile long course through the Mojave desert in no more than 10 hours. On October 8, 2005, 195 teams registered, 23 raced and 5 finished. The robot named “Stanley” finished the course ahead of all other vehicles in 6 h, 53 min, and 58 s, and was declared the winner of the challenge.

In the end of 2006, the version 1.0 was officially released. It was implemented in C and included various image processing operations, computational geometry, face detection, camera calibration, Lucas-Kanade optical flow, motion templates, SIFT features etc. and some classical machine learning methods: decision trees, boosting, SVM, multi-layer perceptrons etc. Also, Adi Shavit created the library logo.

After 2008, OpenCV moved into two new homes – companies named Willow Garage and Itseez. Willow Garage was focused on cutting-edge robotics technology, and Itseez created the best-in-class Computer Vision algorithms. The main people behind OpenCV including Gary Bradski, Vadim Pisarevsky and Viktor Erukhimov join these teams and continue the development of the library.

In 2.0 release, C++ becomes the primary library language. OpenCV also gets automatically generated Python bindings – that is used worldwide since then. Java bindings were added as well.

In 2012, OpenCV went mobile: it started to support first Android, and then iOS. This started the era of on-device Computer Vision, which meant that the algorithms stopped being run at servers exclusively. Mobile OS support enable a whole lot of CV-based applications for smartphones, as well as in-built camera features for computational photography.

Before 2012, the library development was done in SVN, and it was pretty hard for the community to commit their code to the library. In 2012, Kirill Kornyakov and Andrey Kamaev led the migration to Github and created a transparent contribution process that is used up until now. The worldwide community highly endorsed this move, and in 2013, 35 to 50% of all the pull requests that got in OpenCV was authored by the people outside of the core dev team.

In 2015, OpenCV released version 3.0 with major interface improvements. It also included T-API – an OpenCL-based acceleration, as well as a lot of optimizations for platforms like Intel®, AMD® and NVidia®.

In 2016, Intel acquired Itseez. Itseez engineers were the core development team supporting OpenCV. The development of the library returns to Intel – to the same office were it all begain in 1998! However, the core team size has increased by a factor of 3.

Deep Neural Network, or DNN, module has been introduced as a result of GSoC project supervised by Anatoly Baksheev. At Intel, the module has been substantially revised, optimized and expanded to support many popular topologies out-of-the-box. JavaScript interface is another product of GSoC program. Some advanced web technologies, like WebAssembly and WebGPU are used to run OpenCV within a browser efficiently.

Intel releases OpenVINO toolkit for accelerated Computer Vision and Deep Learning on Intel platforms, and it is highly acclaimed by the community for the ease of its use and fantastic inference speed. It has a write-once, deploy-anywhere approach on Intel architectures: CPU, GPU, Movidius VPU and FPGA. OpenVINO can also be used as a backend for OpenCV DNN module – allowing to use OpenCV interface and enjoy significantly accelerated (up to 3x) neural network inference. A generous collection of models named Open Model Zoo is available for free to OpenVINO and OpenCV+OpenVINO users.

In the new major release, the API has been refined to take advantage of newer C++ standard. Graph-API – an efficient image processing engine – was added. The library has been thoroughly optimized for the latest Intel architectures. The new functionality included 3D reconstruction algorithms, QR code detector and more. Intel China contributed Vulkan-based backend for DNN module.

Starting from 2019, the core development team of the library consists of distributed teams at Intel, OpenCV China, and xperience.ai. The core teams and the community work together and expand the functionality and the support of the library.

Recently, we started adding more elements to the ecosystem. Based on our world-class expertise, we created Computer Vision and Deep Learning courses. We also presented OpenCV AI Kit – an innovative chip for Spatial AI that received a very warm welcome from the community. Moreover, we formed a for-profit arm name OpenCV.AI to create products and consult companies on how to create Computer Vision solutions to solve real-life problems.