Caffe layer documentation - Request for feedback

[Jonathan R. Williford]

I have overhauled the Caffe layer documentation. You can find the revised documentation at either of the following two sites:

http://caffe.help/manual/layers.html (requires additional repo, more dynamic)

http://git.caffe.help/tutorial/layers.html (just Caffe repo)

The current (older) version of the Caffe layer documentation is here:

http://caffe.berkeleyvision.org/tutorial/layers.html

I have made a PR to replace the current Caffe layer documentation here:

https://github.com/BVLC/caffe/pull/5148

Please let me know what you think! All of the layers in the Caffe master repository are included. The http://caffe.help site allows for comments (via Disqus) and annnotation via hypothes.is (you should be able to just select text to initiate highlighting or leaving a note).

Cheers,

Jonathan

[Przemek D]

Great work! Immediately lands in my bookmarks, I also gave a thumb up to your PR.

What could be added to the documentation:
* a short "basics" tutorial: what's top and bottom, maybe also a word or two on how to use layer_param field
* param, weight_filler, bias_filler documentation, including allowed filler types and corresponding options (value, std etc.)
* link or description what exactly do the Bias, Scale layers do (or where are they used)?

[Jonathan R. Williford]

Thanks for your feedback (and sorry for the belated response)!

All of your suggestions are good ideas. My PR just got merged and it would be nice to have further PRs to implement your suggestions.

* a short "basics" tutorial: what's top and bottom, maybe also a word or two on how to use layer_param field

Where do you think this should go?

 

* param, weight_filler, bias_filler documentation, including allowed filler types and corresponding options (value, std etc.)

The param is especially a good suggestion. I didn't include the parts of caffe.proto that apply to all parameters (such as lr_mult). I think this should be a separate PR from the weight_filler and bias_filler documentation.

 

* link or description what exactly do the Bias, Scale layers do (or where are they used)?

It would also be good to keep the documentation synchronized with the code documentation when plausible, so it would be nice to update the Doxygen documentation also (I often find the documentation in there confusing).

Evan also had the good idea of promoting the "Layers" documentation so that it isn't just in the tutorials. I think it might be nice to do this sooner rather than later, since this might involve changing the locations of the links.

W dniu poniedziałek, 2 stycznia 2017 20:55:42 UTC+1 użytkownik Jonathan R. Williford napisał:

I have overhauled the Caffe layer documentation. You can find the revised documentation at either of the following two sites:

http://caffe.help/manual/layers.html (requires additional repo, more dynamic)

http://git.caffe.help/tutorial/layers.html (just Caffe repo)

The current (older) version of the Caffe layer documentation is here:

http://caffe.berkeleyvision.org/tutorial/layers.html

I have made a PR to replace the current Caffe layer documentation here:

https://github.com/BVLC/caffe/pull/5148

Please let me know what you think! All of the layers in the Caffe master repository are included. The http://caffe.help site allows for comments (via Disqus) and annnotation via hypothes.is (you should be able to just select text to initiate highlighting or leaving a note).

Cheers,

Jonathan

--
You received this message because you are subscribed to a topic in the Google Groups "Caffe Users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/caffe-users/Hsrj0a19iks/unsubscribe.
To unsubscribe from this group and all its topics, send an email to caffe-users+unsubscribe@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/989b75ce-3e55-491d-b30b-688647743c92%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.