Task board

[Johannes Schönberger]

Hi,

>> I understand that everyone in the core team has a lot of matters to deal with outside the project, but the activity from our (maintainers) side is far less compared to the activity of the contributors. In my opinion, people tend to quit contributing 
> 
> This is true, unfortunately.  How do we increase the number of hours available?  I can see two ways: grow the contributor community (in such a way that contributors know to review as well) or hire someone to work on the project full time.  I'd be interested in exploring the latter avenue.  We can also reach out to employers of existing contributors and try and negotiated dedicated working time on the project.

The latter would be a great step forward!

>> Pt.2 - "issues/PRs management tool for stronger management, not just for its own sake".
>> P.S. I like the practice of `scikit-learn` guys to rename PRs as "[MRG] some awesome contribution" -> [MRG + 1] some awesome contribution". Pretty simple and cheap solution to track LGTM :+1: entities.
> 
> I'd be fine with that.

+1

>> outines just to have a common ground for their functionality. I believe we could still catch the train. So, I propose to consider moving `skimage` infrastructure onto `theano`. That is a hell lot of work, of course, but I'm asking just for a discussion yet.
> 
> We do evaluate these things from time to time.  When GPUs came into fashion, we looked at those, at a recent conference I spoke to folks about using DSLs like Halide for speed optimization, and with our GSoC student Daniil we've spoken a lot about integrating neural nets.  Adopting theano may be quite tricky, so perhaps a proof of concept (or an RFC ;) would be the easiest way to kick off that conversation.

I'd argue that scikit-image wouldn't gain any more traction by transitioning to Theano. The big advantage of theano in the deep learning setting is its automatic differentiation functionality. This will only be useful for a very small subset of functions in scikit-image. The input/output variables to any Python-based deep learning framework are still numpy arrays.

>> From this section, actually, one more pragmatic topic arises. Even if we keep image processing part up to the community, I think that we have to put more our (core team) efforts on the backend part of the project (not just testing and documentation part, but also e.g. make easier function chaining - https://github.com/scikit-image/scikit-image/issues/1529). I'd not expect from anyone outside the core team to have enough knowledge, experience and confidence to implement this and some other features. I.e. I believe that we have to make some necessary contributions to stay competitive.

It would be good, if you could elaborate a little more on this point. Most of skimage's API was designed with exactly that in mind.

Best,
Johannes