It’s often useful to set a fairly low alignment resolution during classification, for particularly heterogeneous datasets. In heterogeneous refinement, one can alter the alignment resolution, without altering the output box size.
In 3D classification, when one sets the target resolution, it alters the box size correspondingly, so if you set a target resolution of say 20Å, you might end up with volumes that are 40px3, even if the corresponding particle sets when reconstructed give a volume of significantly higher resolution.
Would it be possible to add an option to set the output box size separately from that used during classification? This would make inspection of the resulting volumes more convenient.
I’m sure you’re aware this feature is available during 3DVA.
Since 3DVA is imo by far superior I would recommend to use that for serious classifications.
To speed up things you can use binned particles. Just my two cents.
Thanks @tarek - I use 3D-VA for a lot of things, but for some things 3D classification without alignments is still better, when parameters are chosen carefully. In any case they are different approaches, both of which I like to have available! Plenty of serious classifications done using classification without alignments
I just found a case where deep 3D classification without alignments identified several classes that I had not been able to identify using 3D-VA, heterogeneous refinement, or ab initio (or combinations thereof).
Just bumping this - I really like the new improvements to 3D Classification in v4, but I think it would be useful to be able to set the box size used for the final reconstruction (& FSC-based filtering) separately from the target resolution.
