OpenCV team is glad to announce OpenCV 3.0 beta.
The release continues the stabilization process in preparation of 3.0 final, the API is almost rounded, and all the regression tests pass, except for a few tests on 32-bit Windows. We encourage everybody to try it out and definitely appreciate any feedback.
What's new and hot in 3.0 beta?
- The convenient transparent API (nicknamed T-API), where many OpenCV functions can use OpenCL-accelerated path, the technology brought to OpenCV by AMD and Intel companies, continues to evolve. Since 3.0 alpha, performance of OpenCL-accelerated code on Intel Iris Graphics and Intel Iris Pro Graphics has been improved by 10%-230%. That includes faster Canny edge detector, Laplacian, StereoBM, BruteForceMatcher, BackgroundSubtractorMOG2, StitchingWarpers etc. New OpenCL kernels were implemented for several functions: HoughLines, HoughLinesP, cvtColor (selected modes), GEMM, DFT. Thanks to Alexander Alekhin, Alexander Karsakov, Elena Gvozdeva, Vladimir Bystritsky, Sergey Sivolgin and other guys who did that!
- Features 2D, the framework for feature extraction, has been refactored to match OpenCV 3.0 style. Some of the algorithms have been moved to opencv_contrib/xfeatures2d and will be stabilized there.
- About 40 commonly used image processing functions have been optimized using vector NEON instructions, so OpenCV 3.0 beta should run faster, sometimes much faster, on modern ARM chips. Big thanks to Ilya Lavrenov for the amazing job!
- Java bindings have been substantially improved. We now have bindings for the updated features2d, background subtractor classes, a part of opencv_contrib etc. Thanks to Maksim Shabunin for the improvements!
- On x86 and x64 platforms OpenCV binaries (not only pre-built binaries, but also the binaries that you build yourself) include and use a subset of Intel® Integrated Performance Primitives (Intel® IPP) by default. OpenCV 3.0 beta includes a subset of Intel® IPP 8.2.1 with additional optimization for AVX2.
- Over 200 pull requests have been merged in since 3.0 alpha. OpenCV becomes more and more a community-driven project and we are pleased to see that! Big thanks to everybody who contributed the patches to opencv:
edgarriba, Ilya Lavrenov, Alexander Karsakov, Adrien BAK, Rahul Kavi, Vadim Pisarevsky, Maksim Shabunin, Ernest Galbrun, Dmitriy Anisimov, Artur Wieczorek, Alexander Mordvintsev, Alexander Alekhin, vbystricky, Alexander Smorkalov, Rok Mandeljc, Elena Gvozdeva, Marc Rollins, Philipp Hasper, berak, ElenaGvozdeva, PhilLab, Juan Carlos Niebles, Alexander Duda, Alexander Shishkov, xantares, jaco, Brian Park, Chuanbo Weng, Igor Kuzmin, Michele Adduci, Andrey Pavlenko, Bence Magyar, Miroslav Benes, Shubhra Pandit, StevenPuttemans, abidrahmank, f-morozov, micha137, Jose-Luis Blanco-Claraco, Pat O'Keefe, Pavel Vlasov, John Kua, Huu Nguyen, Pierrick Koch, Poly Takahiro Horikawa, GregoryMorse, Felix Endres, Benoit Seguin, jeremy, Tomasz Posłuszny, KruchininD, VBystricky, Dmitry Gribanov, Yan Wang, Yusuke Suzuki, Anatoly Baksheev, bartaandras, Daniel Stonier, dkanafeev, Boaz Stolk, eltermann, Ling Wang@z77x, Manfred Georg, Leszek Swirski, Matthew Conway, Maximilien Cuony, Ariel Elkin
biagio montesano, Vlad Shakhuro, Alex Leontiev, Bellaktris, Vadim Pisarevsky, lluis, dmitriy.anisimov, vludv, Dmitriy Anisimov, Bence Magyar, Daniel Angelov, Alexander Shishkov, ahb, tpietruszka, Maksim Shabunin, Beat Küng, manuele, StevenPuttemans, Alexander Mordvintsev, vbystricky, Philipp Hasper, berak, Sancho McCann, Alexander Alekhin, Ilya Lavrenov, Alexandre Benoit, tolgabirdal, Shuda Li, Rok Mandeljc, VBystricky, Poly Takahiro Horikawa, Michele Adduci, Ernest Galbrun
As usual, the latest version can be downloaded from SourceForge:
- Windows self-extracting archive:
- iOS framework:
- The source code for all the platforms can be downloaded from GitHub: