@stuwilkins in the example notebook of this repo, we find we get zero photons using pycentroids version
'0.2.0rc2'

https://github.com/NSLS-II/centroids/blame/f9a8b54143aab7de6ec0e6dd5f992c93ad33f8ec/examples/Single_Photon_Test.ipynb#L212

Looking at the docstring, it seems that the threshold (250) and filter (3) are inverted for the version deployed on our jupyterserver.

We can update the notebook in a PR, but just wanted to check first.

Also, the histograms look generally the same, but we found 794 events not 702 after just fixing the argument order above. The histograms look largely the same, but the 2x2 scatter plots look like the correlation of fits for xfrac and yfrac are improved. Perhaps this is expected as improvements were made to code. Can you confirm the algorithm should be "better" or do we need to adjust the argument values ?