-
Notifications
You must be signed in to change notification settings - Fork 9
Running RegSeg on Docker
Oscar Esteban edited this page Sep 14, 2017
·
2 revisions
https://www.docker.com/get-docker
docker run -it oesteban/regseg --help
This should give you the following output:
Usage:
General options:
-h [ --help ] show help message
-F [ --fixed-images ] arg fixed image file
-P [ --surface-priors ] arg shape priors
-T [ --surface-target ] arg final shapes to evaluate metric (only
testing purposes)
-M [ --fixed-mask ] arg fixed image mask
-L [ --transform-levels ] arg number of multi-resolution levels for
the transform
-o [ --output-prefix ] arg (=regseg) prefix for output files
-l [ --logfile ] arg log filename
-v [ --monitoring-verbosity ] arg (=1)
verbosity level of intermediate results
monitoring ( 0 = no output; 5 = verbose
)
Optimizer options (by levels):
-a [ --alpha ] arg alpha value in regularization
-b [ --beta ] arg beta value in regularization
-s [ --step-size ] arg step-size value in optimization
-g [ --gradient-scales ] arg alpha value in regularization
-r [ --learning-rate ] arg learning rate to update step size
-i [ --iterations ] arg number of iterations
-w [ --convergence-window ] arg number of iterations of convergence window
-t [ --convergence-thresh ] arg convergence value
--grid-size arg size of control points grid
--grid-spacing arg spacing between control points
-u [ --update-descriptors ] arg frequency (iterations) to update descriptors
of regions (0=no update)
--adaptative-descriptors recomputes descriptors more often at the
beginning of the process
--step-auto guess appropriate step size depending on
first iteration
--convergence-energy disables lazy convergence tracking: instead
of fast computation of the mean norm of the
displacement field, it computes the full
energy functional
Functional options (by levels):
--smoothing arg apply isotropic smoothing filter on target
image, with kernel sigma=S mm.
--smooth-auto apply isotropic smoothing filter on target
image, with automatic computation of kernel
sigma.
--uniform-bg-membership consider last ROI as background and do not
compute descriptors.
-d [ --decile-threshold ] arg set (decile) threshold to consider a computed
gradient as outlier (ranges 0.0-0.5)