As you might know I did something like this with label files for the HCP’s 
multi-modal parcellation, though that was only 360 labels.  If you use the 32k 
mesh and some colors repeat, perhaps it is doable for now?

Peace,

Matt.

From: 
<[email protected]<mailto:[email protected]>>
 on behalf of Timothy Coalson <[email protected]<mailto:[email protected]>>
Date: Wednesday, December 6, 2017 at 5:05 PM
To: Corina Melzer <[email protected]<mailto:[email protected]>>, 
"[email protected]<mailto:[email protected]>" 
<[email protected]<mailto:[email protected]>>
Subject: Re: [HCP-Users] RGBA files in connectome workbench

I just checked the code, and unfortunately there is text matching in a linear 
scan for every label in the label import commands, for the purpose of ensuring 
no duplicate label names, so very large label tables may be quite slow 
currently.  There are possible fixes or workarounds if it is slow enough to be 
a problem.

Tim


On Wed, Dec 6, 2017 at 4:55 PM, Timothy Coalson 
<[email protected]<mailto:[email protected]>> wrote:
Another possibility for a quick and dirty approximation to what you want, you 
may be able to use transparency of layers to blend together coloring from the 
different components, rather than manually mixing them beforehand.  However, 
you would need to play with the palettes, and it couldn't be used directly for 
RGB with full brightness (even if we had the needed palettes).

.label.gii files could theoretically handle 2 billion labels, and our current 
implementation of .dlabel.nii (it uses float32 for all cifti files for 
simplicity) could theoretically handle 16 million without needing to start 
using non-consecutive integers, so these are well over the vertex counts of the 
surfaces (164k per hemisphere).  However, large label tables might 
significantly slow down the coloring or label import code (it uses std::map, 
and I don't think it does any linear scans or text matching for the necessary 
operations, but we haven't tried abusing it this way).

Tim


On Wed, Dec 6, 2017 at 3:57 PM, Corina Melzer 
<[email protected]<mailto:[email protected]>> wrote:
Thanks a lot for your responses. Label files might not be the perfect solution 
for my problem, as the assigned colours are result of a color mixing and are 
more of a continuous than a discrete character. Anyway, I will try to 
approximate this using labels. Is there any kind of limitation regarding the 
number of labels?

Regards,
Corina


On 6. Dec 2017, at 22:00, Harwell, John 
<[email protected]<mailto:[email protected]>> wrote:

Hi Corina,

At this time, Connectome Workbench is not able to display RGBA files.  I do not 
know when, or if, display of RGBA files will be added.

As Matt Glasser indicated in his response, you may want to try using a Label 
file.  A label file contains integers where each value is an index into a table 
of names.

John Harwell

On Dec 6, 2017, at 5:15 AM, Corina Melzer 
<[email protected]<mailto:[email protected]>> wrote:

Hi,

I want to map some custom coloring to the triangles in wb_view and thought of 
using the RGBA files for that. I created this RGBA file with nibabel using 
values between [0,1].

gii_arr = gifti.GiftiDataArray(data=color_data,intent="NIFTI_INTENT_RGB_VECTOR")
gii_img = gifti.GiftiImage()
gii_img.add_gifti_data_array(gii_arr)
gifti.giftiio.write(gii_img, 'test.rgba.gii')

When loading this file into wb_view, all faces are coloured blue instead of the 
defined color codes.

How is this file supposed to be formatted to show the defined coloring? Am I 
getting the idea of this kind of file wrong?

I also tested with an rgb file which has been provided for caret some time ago 
(https://www.nitrc.org/forum/forum.php?offset=125&max_rows=25&style=flat&forum_id=281&thread_id=).
 The overlay is also just blue and not as shown in their provided screenshot.

[workbench version 1.2.3]

Thanks for your help,
Corina









_______________________________________________
HCP-Users mailing list
[email protected]<mailto:[email protected]>
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

_______________________________________________
HCP-Users mailing list
[email protected]
http://lists.humanconnectome.org/mailman/listinfo/hcp-users

Reply via email to