search

timothygrant80 / cisTEM

Other
32 stars 27 forks source link

Better min edge distance in the picker #239

Open arohou opened 4 years ago

arohou commented 4 years ago

I find the current default ends up picking too close to the edges, I often end up ~ doubling the value it chose for me.

bHimes commented 4 years ago

The default looks like it is 128 independent of the particles. It seems like it should at least depend on the particle diameter + delocalization. Ideally it would also consider noise added by electrons that originate outside the detector and are recorded there due to the PSF.

That could get complicated quickly, but if we just used the particle radius + an estimate of delocalization for some common imaging condition that would at least be a step in the right direction. What do you think?

arohou commented 4 years ago

The real min distance should be a function of the box size during extraction. Unfortunately we do not know this at picking time. Perhaps we could set it using the same logic used to propose the default box size later on during refinement package creation? We would have to assume that the user gives the picking job a radius equal to the radius they will declare during ref package creation...