Hi,
In the Patch Motion job, there is an option to “randomly select this many movies to process - helpful for tweaking params”. This is indeed helpful for tweaking params, but the selection is not random - it seems to take a sequential selection of movies, starting from the beginning of the dataset.
A random selection (by default using a fixed random seed) would be better I think, so that one can be confident that the selected stacks are more or less a representative sample of the whole dataset (and not giving misleading results because, for example, there were some crappy squares at the beginning of the dataset).
Cheers
Oli
