Hi,
We prefer to do preprocessing (up to particle extraction) in cryoSPARC-Live running concurrently with the experiment, but curation in the main interface. This is partly due to behavioral differences between streaming and non-streaming 2D, but also the constant and inefficient engagement of GPU resources by streaming jobs.
Would it be possible to refactor 2D classification in the main interface to allow classifying a new, larger particle-stack against templates that were generated at the end of a previous job?
There is an element of use-at-your-own-risk on the part of the operator, but in the scenario I am envisaging, it would be tantamount to skipping the online-EM iterations onto the full-iterations.
Scenario:
2,000 of 5,000 movies collected/preprocessed > Export particles > 2D classification w/ 1 GPU > 5,000 movies collected/preprocessed > Export final particles > 2D classification, full-iterations only against obtained templates w/ 4-GPU parallelisation.
Cheers,
Yang