Removing junk from data


I have a dataset where I am removing junk particles by initially performing ab initio reconstruction (five classes) and then using the output volumes for heterogeneous refinement, again with five classes. The question is do I need to repeat ab initio reconstruction followed by heterogeneous refinement or do I repeat heterogenous refinement with the volumes from the first ab initio reconstruction?

Hi @nooragip,

If your aim is to remove junk particles, the strategy you describe (5-class abinit -> 5 class hetero refine) works well for the first cut. Subsequently, you can the the best classes from the result (just the particles) and use those to start off another 5-class abinit reconstruction. This will start from scratch and further separate particles. So the recommended workflow is:
multiclass abinit -> hetero refine -> select good classes -> multiclass abinit -> hetero refine -> select good classes -> homogeneous refinements of individual classes

Hope this answers your question!

1 Like


Thank you for your clear response, I will try this out with my new dataset. Again, thank you very much.