Compute Feature Sizes and Remove Minimum Size Features 32 bit limit

[imikejackson]

Believe I’ve run into a bug with running Compute Feature Sizes followed by Remove Minimum Size features.

Summary

I have an image stack CT data set (1112 x 1112 x 1762), where I’m segmenting features, calculating the feature sizes, and then deleting the feature sizes beneath a threshold. However, the workflow completely stalls on Remove Minimum Size Features. It won’t crash, it just sits there and never finishes. I tried chunking the data into two separate stacks, and that works. So, I’m pretty confident it is related to the data size.

Couple of observations:

Compute Feature Sizes generates NumElements as an int32.
Shouldn’t this be a uint32 instead? A feature could never be negative in size correct?

• If I multiply the size of my image geometry (1112x1112x1762), this is larger than the largest allowable value of int32. I’m guessing this is the issue
• The pipeline passes the Compute Feature Sizes filter where it is “Validating range of values within input array “Feature Ids”
• My assumption was this would catch any overruns, but perhaps not.
• None of my features are remotely close to the max value of int32 in size, so I don’t think NumElements itself is being maxed out.
• Somewhere within Remove Minimum Size features, I’m guessing it is grabbing the full image stack size, but it is outside the range of acceptable values?

[imikejackson]

This was fixed by PR #1522