Skip to content
This repository was archived by the owner on Aug 5, 2022. It is now read-only.

Conversation

@ipuustin
Copy link
Contributor

Add meta-refkit-extra layer for storing demos and extra content. Have the layer initially in disabled state. See https://lists.yoctoproject.org/pipermail/intel-iot-refkit/2017-March/000005.html for discussion.


# Demo layer, disabled by default. To enable, uncomment the line below
# and increment LCONF_VERSION by one.
# REFKIT_LAYERS += "##OEROOT##/../meta-refkit-extra"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I suspect that this comment might become misleading once bblayers.conf.sample turns into conf/bblayers.conf. A developer editing that bblayers.conf the might uncomment the line and bump LCONF_VERSION, which wouldn't be right.

Do we need this change in bblayers.conf.sample at all?

A developer who wants the extra layer can do "bitbake-layers add-layer ".

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok, I'll remove the comment.

@mythi
Copy link
Contributor

mythi commented Apr 26, 2017 via email

@ipuustin ipuustin force-pushed the meta-refkit-extra branch from 76f2ea5 to fc7c08a Compare April 26, 2017 10:26
@mythi
Copy link
Contributor

mythi commented Apr 26, 2017

@ipuustin will you add #57 content in this PR too?

@ipuustin ipuustin force-pushed the meta-refkit-extra branch from 66effcf to 3af337f Compare April 26, 2017 13:43
@ipuustin
Copy link
Contributor Author

@mythi yes, added now. The only "refkit-affecting" change is the OpenCV Python 3 dependency change.

@ipuustin ipuustin changed the title meta-refkit-extra: add new layer. meta-refkit-extra: add new layer and computer vision demo content there. Apr 26, 2017
@mythi
Copy link
Contributor

mythi commented Apr 27, 2017

@ipuustin thanks, this looks good to me. Let's take it once the documentation is finished.

@ipuustin ipuustin force-pushed the meta-refkit-extra branch 2 times, most recently from a00ab01 to 52f2ac1 Compare April 27, 2017 12:28
@ipuustin
Copy link
Contributor Author

Ok, rebased. The docs are now in ok shape.

Copy link
Contributor

@pohly pohly left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So "refkit-extra.conf" is not meant to be a distro configuration? Currently it is, because it is placed in conf/distro of a layer.

Better move it to conf/distro/include. I would also squash the two commits introducing meta-refkit-extra.

ipuustin added 10 commits April 27, 2017 19:21
Add meta-refkit-extra layer for storing demos and extra content. Have
the layer initially in disabled state. Also add initial configuration.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
Pillow is a fork of PIL, Python Imaging Library.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
ImageIO is a Python library for doing basic I/O operations for images.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
The protobuf package didn't compile Python bindings. Add a new package
which does that.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
The goal is to have the computer vision stack to be scriptable by
Python 3.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
The recipes are originally from (non-maintained) meta-convnet layer.

Patch Caffe to compile the Python bindings with Python 3. Also change
the image IO libary away from scipy-image, so that we don't have full
matplotlib and graphical environment as runtime dependencies. Then add
custom demo scripts.

Many recipes are updated and customized. They probably don't work
properly in anything else than 64-bit x86 environment.

Also add package with a pretrained imagenet model. The model is a
pretrained deep learning network for classifying based on the imagenet
database.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
Pyrealsense contains Python bindings for librealsense.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
Added documentation for setting up pyrealsense and Caffe demos.

Signed-off-by: Ismo Puustinen <ismo.puustinen@intel.com>
@ipuustin ipuustin force-pushed the meta-refkit-extra branch from 52f2ac1 to 2015b44 Compare April 27, 2017 16:26
@ipuustin
Copy link
Contributor Author

@pohly : ok, I did the changes you suggested.

@@ -0,0 +1 @@
PACKAGECONFIG += "python"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ipuustin Is this needed? IIRC you mentioned it's not needed.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's still needed -- it enables Python support in boost. We no longer need to select Python 3 by editing the user-config.jam file, because Python 3 support is now enabled as default if it's found. See difference to the previous iteration of the same file. https://github.com/ipuustin/intel-iot-refkit/blob/165b46149a2c5a58788b91d2cac97c394e567f14/meta-refkit/recipes-convnet/boost/boost_%25.bbappend

However, it's a good question if we should move the PACKAGECONFIG variable to the layer distro configuration file instead of the .bbappend file. What do you think?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, I see. python isn't enabled by default. I got confused by what you wrote earlier: "The bbappend should not be needed anymore."

@mythi mythi merged commit 82a1b90 into intel:master Apr 28, 2017
@pohly
Copy link
Contributor

pohly commented Apr 28, 2017 via email

@mythi
Copy link
Contributor

mythi commented Apr 28, 2017 via email

@pohly
Copy link
Contributor

pohly commented Apr 28, 2017 via email

@mythi
Copy link
Contributor

mythi commented Apr 28, 2017 via email

@ipuustin
Copy link
Contributor Author

I think @mythi means that this does not matter much because in order to properly use the layer, the user needs to do two things: add the layer and include the configuration file. This will anyway cause the build result to change.

I like the concept of separating policy (configuration files) and mechanism (recipes), even though it makes recipes-convnet directory not to be "self-contained" anymore. I can do a PR which moves the PACKAGECONFIG to the configuration file.

@ipuustin
Copy link
Contributor Author

PR here: #133

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants