Jetson/Tutorials/OpenCV

From eLinux.org
Jump to: navigation, search

The easiest way to start using OpenCV is to download the prebuilt .deb packages for CUDA and OpenCV for Linux4Tegra (L4T) that is the default OS of the Jetson TK1 board.

  • First, open a console/shell/terminal on your device and make sure you enabled universal repositories:
sudo apt-add-repository universe
sudo apt-get update
  • Now install the CUDA toolkit onto the device. Follow all the instructions for Installing CUDA then return back here when CUDA is working.
  • Download the 2 OpenCV L4T prebuilt packages from the Jetson TK1 Support Page. Then you can simply install them onto your device:
sudo dpkg -i libopencv4tegra_2.4.8.2_armhf.deb
sudo dpkg -i libopencv4tegra-dev_2.4.8.2_armhf.deb

You should now be able to build & run OpenCV applications natively on your Jetson TK1 device, including with the CUDA-optimized "gpu" module (that works exactly the same as the OpenCV "gpu" module on a desktop computer) as well as the Tegra CPU-optimized (NEON SIMD + Multi-threading) speedups due to OpenCV4Tegra (these optimizations are done automatically for you and thus you don't need to enable them in your OpenCV code).

More info

Now that CUDA & Tegra optimized OpenCV is installed on your device, you can build & run some of the OpenCV sample CPU & GPU programs as mentioned in the Testing OpenCV section of the Installing OpenCV reference page. That reference page also shows how to get the OpenCV documentation and official tutorials, and how to build the OpenCV library from scratch such as if you want to modify the source code.

If you want to see the typical power draw for various OpenCV and computer vision applications, or to learn more about how to use the CUDA "gpu" module in OpenCV or more about OpenCV4Tegra, read them at the OpenCV Performance page.