Background MIPs for TFE

[chantelleleveille]

Add backdrop images to TFE

Generatating MIPS

• Raw images are loaded from S3 (quilt)
• Maximum Z projection images are generated the same way as Figure 1 image.
• Image is downsampled by half to make it load very quickly on the website.

Note: When I made the max intensity z projects using skimage resize method the resulting image looks much better than the precomputed zarr resolution_level equivalent. I made it an option in the function to do it either way, but the default behavior is im1! See example below:

im1 = save_colony_backdrop_mips('small', figdir, downsample_factor=2, zarr_res_level=0)
im2 = save_colony_backdrop_mips('small', figdir, downsample_factor=None, zarr_res_level=1)

Skimage resize (im1)	Precomputed Zarr (im2)

The difference is that the precomputed zarr uses skimage resize with the order of interpolation set to nearest neighbor, whereas the skimage resize default in this method is set to bi-linear!

TFE integration

• Backdrops only need to be generated once. This step will be slow.
• After that use the flag --make_backdrops False to use previously generated images (similar to --noframes)

Here is a preview link for the Large colony in the exploratory dataset. Preview Link (will only work internally)

[ShrimpCryptid]

LGTM! Excited to see this change in the NucMorph datasets :)

[jcass11]

Testing by running with default behavior for one colony and filter/dataset (small colony, exploratory) pdm run nuc_morph_analysis/lib/visualization/write_data_for_colorizer.py --dataset small --filter exploratory_dataset
I will follow up when it finishes!
I looked at the sample link and made sure the features look right/there's no flickering on segmentations or out of order frames (common accidents that creep in to TFE data) and generally played around and everything looked great/seamless
Beautiful work!

[jcass11]

My test of this update ran without error and everything looks good! Thank you!

[chantelleleveille]

@pgarrison @cfrick13 this looks like it has 2 approvers but I still need one more because Peyton's doesn't count.

[ShrimpCryptid]

@pgarrison @cfrick13 this looks like it has 2 approvers but I still need one more because Peyton's doesn't count.

😔