Announcement

Collapse
No announcement yet.

OpenCV 4.0 Released As The Overhauled Computer Vision Library, Adds Experimental Vulkan

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • OpenCV 4.0 Released As The Overhauled Computer Vision Library, Adds Experimental Vulkan

    Phoronix: OpenCV 4.0 Released As The Overhauled Computer Vision Library, Adds Experimental Vulkan

    OpenCV 4.0 is now officially out as the widely-used real-time computer vision library...

    http://www.phoronix.com/scan.php?pag...V-4.0-Released

  • #2
    YOLOv3

    Comment


    • #3
      AFAICS, this is still the RC. At least, that's what their release webpage lists.

      Comment


      • #4
        Originally posted by GruenSein View Post
        AFAICS, this is still the RC. At least, that's what their release webpage lists.
        We are glad to announce OpenCV 4.0 release. https://github.com/opencv/opencv/releases

        Comment


        • #5
          Originally posted by TomPollok View Post

          We are glad to announce OpenCV 4.0 release. https://github.com/opencv/opencv/releases
          On the main webpage opencv.org, there is no final announcement.

          Comment


          • #6
            These are some pretty good changes. It's nice that they focus on both CPU and GPU performance.

            Comment


            • #7
              Originally posted by GruenSein View Post

              On the main webpage opencv.org, there is no final announcement.

              Comment


              • #8
                Originally posted by GruenSein View Post

                On the main webpage opencv.org, there is no final announcement.
                There you go: https://opencv.org/opencv-4-0-0.html

                Comment


                • #9
                  Originally posted by schmidtbag View Post
                  These are some pretty good changes. It's nice that they focus on both CPU and GPU performance.
                  I am actually surprised, how diverse their acceleration efforts are. There is basically every vectorization instruction set and best of all: It is selected at runtime. No need to recompile for each deployment target. That makes it so much easier to use the available hardware regardless of the node you are on. Then, there are many parallelization schemes (OpenMP, IPP, TBB). And beyond this, hardware acceleration is available via OpenCL and now Vulkan. Quite nice to see that. That must have been a crazy amount of work.

                  Comment


                  • #10
                    Originally posted by GruenSein View Post
                    I am actually surprised, how diverse their acceleration efforts are. There is basically every vectorization instruction set and best of all: It is selected at runtime. No need to recompile for each deployment target. That makes it so much easier to use the available hardware regardless of the node you are on. Then, there are many parallelization schemes (OpenMP, IPP, TBB). And beyond this, hardware acceleration is available via OpenCL and now Vulkan. Quite nice to see that. That must have been a crazy amount of work.
                    I totally agree. And yes, I'm sure it is a lot of work, but, these are the sorts of things you really only have to worry about once.
                    OpenCV is one of very few libraries that really needs as much optimization and performance gains as possible, since it is often used for robots and other small/portable devices with limited compute power. So, the fact that the devs have such a wide variety of options is very forward-thinking, and makes this one of the most well-planned open-source projects out there (in my opinion, anyway).

                    Comment

                    Working...
                    X