Introduction
This article documents how to install VisionPack into a Yocto-based distribution using the provided meta-deepview package. The packages within VisionPack come in binary form which should support most recent versions of Yocto, recent being circa 2018 or newer, as we strive for good binary ABI compatibility.
Note: packages supporting musl, uclibc, or other custom libraries can be made available upon request. Please submit a ticket with your specific requirements if the provided libraries do not meet them.
Requirements
- ARM64 target architecture.
- VisionPack is optimized for the i.MX8M Plus SOC.
- Yocto tested with Dunfell and Kirkstone.
- Python 3.6 and GStreamer 1.0 are optional.
- Build machine suitable for building a Yocto distribution.
Instructions
Add the meta-deepview package to either your repo manifest.xml or manually git clone the repository into your Yocto sources (sometimes called layers) folder.
Add an entry to your conf/bblayers.conf such as the following, adjust depending on how you've organized your build layout.
BBLAYERS += "${BSPDIR}/sources/meta-deepview"
The quickest way to add VisionPack to your custom image is to add "visionpack" to your IMAGE_INSTALL, such as by adding this line to your conf/local.conf.
IMAGE_INSTALL:append = "visionpack"
The meta-deepview provides bbappend files for the NXP Yocto imx-image-multimedia and imx-image-full images which add the visionpack packagroup which means the bblayers.conf edit is all that is needed when using these images.
Refer to recipes-visionpack/packagegroups/visionpack.bb for details of the packages installed. In short, we install Deep View RT, VAAL, and VideoStream including their respective GStreamer and Python extensions.
Usage
With VisionPack integrated into your Yocto distribution you will now be able to run the various reference code examples and any of the downloads from our AI Application Zoo (requires a display, contact us for headless or remote streaming options).
Comments
0 comments
Please sign in to leave a comment.