Display multiple arrays with translation or other transform applied?

[dpshepherd]

Related to #60 , I was curious if there are plans (or maybe already possible and I missed it) to support displaying N array objects with known translation, or more complicated spatial transforms, relative to each other.

The motivation is that I am trying to finish up some documentation for our GPU-accelerated 3D merfish analysis package. As part of the documentation, we have a few examples that analyze publicly available data. Often, these data have unclear rotations, flips, and reversal w.r.t. the global coordinate system that the stage reports. The only way to figure this out is to manually inspect the tiles and look for features that should overlap.

Right now, we spin up a napari instance and figure out the correct orientation for a few tiles. Ultimately, we programmatically include these transforms and then use multiview-stitcher to calculate the rigid transforms to place all of the tiles into the stage's global coordinate system.

It would be great to use ndv instead of napari, because we have worked hard to reduce dependence on certain packages to streamline performance.

[gselzer]

Yes, there are definitely plans to display multiple arrays.

This is a good issue, as while both the discussions in #60 and also in #81 require the ability to display multiple datasets, we hadn't talked about a full (linear) transform before, which would be a useful and (relatively) easy feature once we can actually reason about multiple datasets.

[tlambert03]

hey @dpshepherd, nice to see you here as always :)
yep, echoing @gselzer, this is definitely in scope. reducing the dependencies and latency to display datasets is right at the top of the motivating factors here, and we do want to support placing multiple datasets in a global coordinate system. Thanks for opening this issue and stay tuned

[dpshepherd]

Thanks @gselzer and @tlambert03! You always seem to be working on things quite useful to us!

@tlambert03, this merfish library has been eating all my free time. I had to unexpectedly take it over, but it's stable enough for usage on big datasets now. I'm hoping to get back to helping the group with microscope control once I finish cleaning up docs and tests.