Teaching lesson in disciplines where single-channel data is the norm

[ivastar]

In astronomy scientists use single channel CCD images for their research. Color images are only created for publication purposes and for public outreach. There is a lot of very useful content in this lesson that can still be taught for single channel images but maybe some thinking needs to be done on how to present the material to learners who have rarely/never encountered multi-channel/color images in their work.

One suggestion that came from a discussion at CarpentryCon is to use separate images at different wavelengths as the different channels. Images need to be registered in order for this to work.

Are there other disciplines that use primarily single-channel data?

[bobturneruk]

Lots of optical microscopy (and I think also some satellite imaging). Often these are quite well registered (i.e. aligned across channels), but that's another consideration that I hadn't thought of.

Does including RGB as standard place too much of a burden on learners only interested in single channel data, I wonder? And related, can learners go on to apply knowledge of RGB to deal with different types of multi channel data e.g. two infra-red bands? Maybe some more "information boxes" in the initial episodes would be sufficiently helpful?

[K-Meech]

Just to add that this is also common with scientists who mostly work with electron microscopy, or X-ray microscopy data. There are rarely RGB images.

Some extra 'information boxes' are a good idea. It would be nice to make it clearer how these principles of working with RGB can be extended to any multi-channel data (which is also common e.g. with multi-channel fluorescence microscopy images)

[tobyhodges]

@ivastar I was chatting with @tischi earlier this week and the prospect was raised of a hackathon-like event to try sketching out alternative versions, forked from this lesson and modified to use different central example datasets. We discussed the idea of one version using the kind of microscopy data mentioned by @K-Meech above (@tischi expects to know plenty of people who might contribute to that effort), and a second for the kind of astronomical data you mentioned above.

It sounded like there was some potential for such an event to be hosted in Heidelberg. Does this sound interesting to you?

[mkcor]

It sounded like there was some potential for such an event to be hosted in Heidelberg. Does this sound interesting to you?

If I may chime in, I don't live very far from Heidelberg. If you don't have a full team of instructors/helpers yet, I could join perhaps... depending on when you schedule the event (e.g., I'm not available in March).

[marcodallavecchia]

I want to go back to this issue as several discussions have finally matured into a concrete initiative driven as a collaboration between The Carpentries and GloBIAS to create a parallel curriculum on Bioimage analysis with Python focused on relevant bio-image fluorescence microscopy data. It's currently just getting started here: https://github.com/carpentries-incubator/bioimage-analysis-python

During those discussions, we reached a consensus in that different fields of expertise could create other similar parallel tracks to tackle more niche-specific topics not covered in this more general ones. So while the Bioimage analysis with Python curriculum will not cover astronomy data, it will definitely cover much less natural images and focus on multidimensional images (multiple channels, time, Z slices etc..).

Based on this, we think this issue will partially be addressed in this way via parallel curriculum, and we will keep the focus on natural images in this one.