search

yuki-biology / nuclei-volume-calc

0 stars 0 forks source link

Possible Volume Estimation Performance Improvement #1

Open william-silversmith opened 2 years ago

william-silversmith commented 2 years ago

Hi there!

I was just browsing through some repos that are using cc3d and I saw that you might benefit from some performance improvements. I am not sure how crucial this is to your project, but the following code can be made much faster.

https://github.com/yuki-biology/nuclei-volume-calc/blob/23b9362c39ef8507ed6b138d55e1534b2d6ccb2c/module/cc3d.py#L23-L28

This code requires repeated calculation on the image and has bilinear performance. You can get the same result much faster by using cc3d.statistics

cc_out = cc3d.connected_components(labels_out, connectivity=connectivity)
stats = cc3d.statistics(cc_out) # calculates centroids, bboxes, and voxel counts in one pass
voxels_per_label = stats["voxel_counts"]  # ndarray, index is CCL ID

Thanks for reading and I hope I'm not intruding. Good luck on your project!

Will

yuki-biology commented 2 years ago

Thank you for your advice! I will try to fix as you said.