classesbymra

This program classifies a set of particles based on a set of references (usually projections).

classesbymra <particles> <refrences> [split] [frac=<num>/<denom>] [verbose[=<n>]] [mask=<rad>] [refmask] [imask=<rad>] [maxshift=<rad>] [precen] [sigfilt] [tree1=<treefile>] [tree2=<treefile>] [ctfc] [matrix] [sep=<nspl>] [logit=<label>] [rfp] [norot] [usefilt] [refine] [slow]

Parameters:


<particles>Raw particle filespec
<refrences>Reference image filespec
[split]Always use this
[frac=<num>/<denom>]Operate only on a fraction of the raw images
[verbose[=<n>]]More verbose output
[mask=<rad>]Mask the classified particles, also reduces noise for better classification
[refmask]This will use a mask generated from the reference image. Assumes the 3D model has been sensibly masked.
[imask=<rad>]Inside mask, classified particles are masked for classification, but not in the cls files.
[maxshift=<rad>]Maximum translation during image alignment
[precen]Particles are assumed to be accurately precentered
[sigfilt]Apply a real space filter so only high amplitude data is used for classification
[tree1=<treefile>]First stage high speed classification, see <a href=projtree.html>projtree</a>
[tree2=<treefile>]Second stage high speed classification
[ctfc]Projections are CTF corrected before comparison
[matrix]Generates similarity matrix for entire data set in matrix.dat
[sep=<nspl>]Puts each raw particle in 'nspl' different classes.
[logit=<label>]This specifies a label to use in the particle.log file. This records particle number classification throughout a refinement.
[rfp]This will classify using 'rotational footprints' instead of the images themselves. Much faster and quite accruate.
[norot]This will not do any rotational alignment in classification. Primarily used with projections and the matrix option, but may have other uses.
[usefilt]Experimental. Use the 'filt' version of the input file for classification.
[refine]Refines 2D alignment done with fast algorithm. Slower, but not 'slow'.
[slow]Uses a very slow, exhaustive search algorithm for alignment

Usage:

classesbymra start.hed proj.hed split mask=28 ctfc sep=2 tree1=proj.tree

Description

This classifies a set of raw particles based on a set of reference projections. The output is a set of cls*.img files, one file for each reference image. If the projections were generated from a CTF corrected model, specify the 'ctfc' option so the projections will be 'decorrected' before classification.

The 'tree' options implement 2 part classification for higher speed. Check projtree details on this.

The 'sep' option allows each particle to be placed into several different classes. If the images are very noisy, or if the angular spacing is small, misclassification of some particles will inevitably occur. The sep option makes it very likely that one of the classes each particle ends up in will be the correct one. The classalignall procedure can then be used to obtain sufficient consistency within the set of images actually used in each class average.


EMAN Manual page, generated Tue Oct 8 21:38:47 2002