OpenCV 5.1 Brings More AVX-512 Optimizations, Other Improvements

OpenCV 5.0 debuted back in June with a rewritten DNN engine and many other exciting improvements for this widely-used computer vision (CV) library. Out today is OpenCV 5.1 with tons of fixes and other incremental improvements to this important code. OpenCV 5.1 brings a lot of changes for advancing computer vision in the era of AI. Some of the notable changes catching my attention with OpenCV 5.1 include: - New AVX-512 optimizations along with various other SIMD optimizations. There are also some AVX-VNNI improvements too. - Expanded FP8 support in different areas. - The classic DNN engine was removed in favor of keeping just the new engine, which debuted in OpenCV 5.0. - NVIDIA CUDA and cuDNN improvements to the DNN module and into other areas of OpenCV. - A number of new RISC-V RVV vector kernels. - An OpenGL-based 3D visualizer is added for meshes, point clouds, RGB-D frames, and camera trajectories. - Improved bindings for Java, JavaScript, and Python. - Hundreds of bug fixes. OpenCV 5.1.0 is available from GitHub.