Rendering Large X/Y/(RGB) In N5 Viewer

[jmdelahanty]

Hello N5 Devs! I was curious if there was any help I could find for rendering an image that is X/Y/(RGB) in the N5 viewer. I have an HDF5 dataset that's about 10GB. Its shape when inspected using:

import h5py

with h5py.File('test.hdf5', 'r') as f:

    dataset = f['test'] 

    print(dataset.shape)

Gives me: (21534, 88200, 3)

When I open it in N5 (it opens just fine), I don't actually see my image for some reason! It is displayed as in the viewer as this:

test (3 x 88200 x 21534)

And all I get is a big blank box:

I've been having some trouble getting it to view my big 2D image properly, any tips on how I can get it working? Thank you!

[jmdelahanty]

This actually appears to function correctly, it can just be really tough to find your image after it's been loaded. Rotating around a little in the viewer finds a sliver of my image in the space it makes that I can then zoom into and, eventually, visualize well! Interesting... but since I'm looking at a 2D image I would think I should be able to stay around my the image easier. Am I missing something?

Also your lab is super cool and I love these tools.

[bogovicj]

Thanks @jmdelahanty for reporting this weird behavior and nice words ❤️ .

My guess is that n5-viewer made some incorrect assumptions about the data (possibly that it's 3D, not 2D+color).
I'll try to replicate with an h5 and I expect it will be easy to do so with the information you've given.

Long term, I'd like to provide users with an easy way to assign meaning to the the dimensions of their data. There is currently a semi-programmatic way to do this which I can explain in a follow up message after I replicate.

Medium term - would you be open to writing a little extra metadata in your h5's that will tell n5-viewer how to interpret the data dimensions? If so, I can give you an example.

[jmdelahanty]

I love metadata! More than happy to include some in the file. Thank you for your quick response and help! I'm really grateful!