Add OpenCV to STM32MP157C-DK2 Discovery Board

[DRuss.2]

I'm trying to figure out how to add OpenCV to my STM32MP157-DK2 Discovery Board. I am not new to ST Microcontrollers, but I am new to the MP1 and Cortex A processors. I'm also not new to Linux but am completely new to embedded Linux and Yocto.

I can see that there is a recipe for OpenCV included in the Ecosystem V1.2 release distribution package, but when I build the release and flash, OpenCV is not present on the DK2 board. I was able build OpenCV specifically with the bitbake command, but again, it is not present in the image. I have seen a few references to installing OpenCV in the wiki but not enough details to get to the next step.

I am wondering what the correct or preferred method for getting OpenCV included in the image. Preferably, enough information or pointers to documentation that I can understand the bigger picture and know how to do this in general and not simply how just to add OpenCV in this case.

[Christophe Guibout]

Hello @Community member​ ,

If you want to add a recipe into your image, you need to add this recipe in the image recipe (st-image-weston.bb if you build this image)

Feel free to look into the wiki.

BR,

Christophe

[mleo]

Hello @Community member​ ,

you can also use apt-get mechanism to install on board required packages/libraries.

See article https://wiki.st.com/stm32mpu/wiki/Package_repository_for_OpenSTLinux_distribution

Board $> apt-get update

Board $> apt-cache search opencv

I believe you will find the one you need.

BR,

Milan

[wzhan.3]

Hello DRuss, I want to know if you have solved this question. cause I also need to add OpenCV to my MP1 board. Look forwward your reply!

[DRuss.2]

I abandoned developing for the MP1 board long ago, so there is not much I remember or can help you with. The answer I ultimately came up with was to add my own layer to yocto and rebuild. The layer brought in OpenCV. I'll paste my notes below, but I have to admit that they don't make much sense to me now. If they don't make sense to you then, I apologize, but there is not much I can do at this point. I know this worked and that we were able to run OpenCV, but we did not get the performance that we needed and moved onto other things (and other projects).

- https://wiki.st.com/stm32mpu/wiki/How_to_create_your_own_image

- https://wiki.st.com/stm32mpu/wiki/How_to_create_a_new_open_embedded_layer

- First, create a new layer.

 - source your BitBake enve setup

  - cd STM32MPU_workspace/STM32MP15-Ecosystem-v1.2.0/Distribution_Package/openstlinux-20-02-19

  - source ./layers/meta-st/scripts/envsetup.sh

 - Get a list of existing layers

  - bitbake-layers show-layers

 - Create the new layer

  - Specify:

   - Location of the new layer

   - Name of the new layer

   - (Optionally) the priority of the new layer

  - bitbake-layers create-layer ../layers/meta-foo/meta-foo-image

- Add the new layer to your configuration

 - bitbake-layers add-layer../layers/meta-foo/meta-foo-image

- Add a recipe for the image

 - cd ../layers/meta-foo/meta-foo-images

 - mkdir recipes-foo

 - cd recipes-foo

 - mkdir images

 - cd images

 - touch foo-image-weston.bb

 - gedit foo-image-weston.bb

 - Add the following to the file

   require recipes-st/images/st-image-weston.bb

   SUMMARY = "Image modified for FOO"

   #

   # INSTALL addons

   # Note - Yocto documentation says that using CORE_IMAGE_EXTRA_INSTALL here

   #    is incorrect and that this variable should only be used in local.conf.

   #

   CORE_IMAGE_EXTRA_INSTALL += " \

     opencv     \

   "

 - Save and exit gedit

- Build Image

 - bitbake foo-image-weston

- Flash Image

 - Out of Scope for this document

- Generate SDK

 - bitbake -c populate_sdk foo-image-weston

- Generate Extended SDK

 - bitbake -c populate_sdk_ext foo-image-weston