3DFlex reconstruction and masking

[yoshiokc]

Where does 3DFlex pull the mask used for the reconstruction job? And is there a way to ‘trick’ it to use a different mask? The reason I ask is that I think I’m seeing overfitting in my 3dflex recon job (for this particular dataset) and when I look at the log my guess is that it is upsampling the hard mask used in defining the original mesh model, and this mask falloff is far too abrupt in this particular circumstance. Would love to be able to supply a different mask for the recon job that is separate from the mask used to define the original mesh model.

Thanks,
-Craig

[yoshiokc]

Noting here that increasing the number of BGFS iterations to 60 helped produce a more “normal” looking FSC, but I think the mask looks a lot sharper than I’d normally like.

[apunjani]

Hi @yoshiokc,
Thanks for posting - you may already have figured this out but to confirm, in the Flex Reconstruct job, the “Masked” FSC you see in this plot is not computed with the hard mask, but rather the soft mask that was created in the upstream Flex Mesh Prep job. We’ve noted to add a plot of the soft mask in the reconstruction job to make this more clear.
Thanks!