napari for geodata

[jni]

🚀 Feature

We've had some interest in using napari for geodata. In addition to physical coordinate systems, we'll need tiling and geographic transforms. It turns out there is a project implementing these things, which I heard about on gitter.im/vispy/vispy: https://github.com/codypiersall/vismap

I have nothing more to add here, only opening this to track interest and progress here.

[sofroniewn]

Very cool, maybe @freeman-lab can help us out ;-)

[GenevieveBuckley]

Is this tiling purely for map data, or could you also tile images with the same mechanics?

[jni]

That specific package is for map data, but yes, it certainly can and should be generalised when it comes to napari.

[EmanuelCastanho]

I'm trying to use GeoTiffs and Shapefiles in napari and I consider this feature very important to create a 3D visualization that facilitates the analysis of certain parameters over time. Can you provide more information on how to implement vismap in napari and how to add a projection (example EPSG 4326)?

This is what I've achieved so far:

• Added a shapefile (it is upside down), represented by several points in napari;

• 3D visualization (Lon, Lat, Time) of the NDVI (vegetation parameter) values of an island, important to detect clusters.

Without a geographic coordinate system it is difficult to use this type of data, any tips? I would like to help develop this feature for napari. I think it would be interesting to achieve something similar to this (but simpler and open source): https://www.esa.int/ESA_Multimedia/Videos/2018/07/Earth_System_Data_Lab/

[GenevieveBuckley]

This is great! More geodata examples would be wonderful. I have started a "napari for geoscience" playlist on youtube, which currently has only Draga's talk in it, but ideally we can add more in the future: https://www.youtube.com/channel/UCbTgw84ew4pxTJ9qu3W2hqg/playlists

[jni]

@EmanuelCastanho awesome. I'd love to do a bit of pair programming with you to help you get to where you're going! Bonus points for repeating the above analysis on Tabuaeran! :smiley: Anyway you can book a slot with me at: https://twitter.com/jnuneziglesias/status/1385754529639985152 I think @DragaDoncila will join us. :smiley:

For geographic transforms, we'll probably need to do a bit of an overhaul of our transform system to enable user- or plugin-defined transforms, but I'm hoping it's not too terrible. :grimacing:

[sofroniewn]

Super cool you're excited about napari for geosciences!

For geographic transforms, we'll probably need to do a bit of an overhaul of our transform system to enable user- or plugin-defined transforms, but I'm hoping it's not too terrible. 😬

I'd be interested to know what this requires too, I'm not so familiar with geographic transforms. cc @andy-sweet who was worked on the napari transform system too

[DragaDoncila]

Definitely excited for this and happy to contribute to napari and/or any plugins we need to get this work across.

[andy-sweet]

I'm also unfamiliar with geographic transforms, but happy to learn about them. In the architecture working group, we split off a topic on potentially revamping transforms in napari, so I added this issue to the list.

In general, it feels like transforms may be coupled with the layer data slicing problem (e.g. if we want to handle tiled data). Currently, I think napari's model is mostly about getting all the data in memory then providing vispy with the forwards matrix transform (though I think that's a bit of a simplification, because we do have Layer._transforms['tile2data']). But I have been thinking about whether we want to invert that model - i.e. mostly think about the viewer/canvas requesting the data required for some field of view. It would be good to understand how much existing packages (like vismap and others) can help us to implement something like that.

[sofroniewn]

But I have been thinking about whether we want to invert that model - i.e. mostly think about the viewer/canvas requesting the data required for some field of view.

@perlman might be interest in that approach too as we think about the "level of detail" rendering - cc @nclack

[GenevieveBuckley]

Just adding a note here to say that Vlad is maybe interested in writing a GeoJSON writer plugin for napari https://forum.image.sc/t/annotation-file-specification/58910/2 (it doesn't look like Vlad specifically uses geodata, but a plugin for this would obviously be useful in lots of situations, so I think it fits this thread too)