Currently, MemBrain-pick assumes that the rlnCoordinate<X/Y/Z> columns are always in (non-centered) Å. This is not compatible with RELION-5 assumptions though.
RELION-5 accepts input coordinates in one of two ways:
rlnCoordinate<X/Y/Z> in non-centered, unbinned pixelsrlnCenteredCoordinate<X/Y/Z>Angst in centered Å
Currently, we ask users in the documentation to apply a scaling factor in order to meet 1) correctly, i.e., convert back to pixels. This is a bit clumsy. A better way would be to use the pixel size again before generating the output of the predict step, to convert the coordinates back to pixels in the way RELION expects in 1).
Currently, MemBrain-pick assumes that the
rlnCoordinate<X/Y/Z>columns are always in (non-centered) Å. This is not compatible with RELION-5 assumptions though.RELION-5 accepts input coordinates in one of two ways:
rlnCoordinate<X/Y/Z>in non-centered, unbinned pixelsrlnCenteredCoordinate<X/Y/Z>Angstin centered ÅCurrently, we ask users in the documentation to apply a scaling factor in order to meet 1) correctly, i.e., convert back to pixels. This is a bit clumsy. A better way would be to use the pixel size again before generating the output of the predict step, to convert the coordinates back to pixels in the way RELION expects in 1).