seeing this utility, I would also like to have the option to disable randomization during heterogeneous classification to allow a quality subset of particles to “seed” the refinement (along with the reference bias) and establishment of euler distribution when also applying to multiple-millions of particles. Which reminds me that a “streaming heterogeneous refinement” would be extremely useful as well. Streaming heterogeneous refinement (feature request)
The references that work well for 1 million particles generally work just as well for 10 million or for 10 batches of 1 million. Would be great to link the streaming 2D particles (maybe even just those which are selected) into a continuing hetereogeneous refinement so I don’t have to export and run particle sets over and over (or wait until collection is complete to start het refine of multiple millions).