Hi,
We've found and fixed a problem with the multichannel priors
init routine, binaries will be in the next patch, but here's the
replacement file to build a working binary:
Regards
Matthew
> On 28.11.2007, at 17:55, Matthew Webster wrote:
>
>> There appears to be a problem with using priors (the -a
>> identity.mat option) on your data set. A test run with priors
>> disabled produces output with multiple labels...
>
> Yes, exactly. It's only the combination of priors and multiple
> channels that has problems.
>
> Unfortunately we are specifically interested in using both, because
> without priors the labeling has been very inconsistent across our
> data sets and with only T1 the segmentation has capabilities of
> improvement.
>
>
> Regards,
> Timm Wetzel
|