[Neuroimaging] [PySurfer] Brain's save_image method produces images with only background color
Gael Varoquaux
gael.varoquaux at normalesup.org
Fri Sep 16 12:29:40 EDT 2016
Nx has always been a problem with Mayavi (or actually VTK, which is the
underlying technology). Basically, it interfers with the openGL contexts,
and in some cases the buffer cannot be captured well. Hence the black
image.
IMHO, the bug is in NX or the mesa driver, or both.
Gaël
On Fri, Sep 16, 2016 at 12:00:15PM -0400, John Pellman wrote:
> Pysurfer isn't running headless- it's using x2go, which is based upon the nx
> protocol, a technology that improves the ability of X11 to function over a
> network connection. Therefore, I don't think that Xvfb is related. xvfbwrapper
> might be usable as a workaround, however.
> As I mentioned in my last post, I traced the offending method back to mayavi.
> I've opened an issue related to this here.
> Kirstie- if you'd be willing to refer your sysadmin to this thread I think that
> would be great, as I would be interested in hearing what theories or potential
> fixes he/she might have for this issue as well.
> --John
> On Fri, Sep 16, 2016 at 5:43 AM, JB Poline <jbpoline at gmail.com> wrote:
> That's a cool idea and package - thanks for pointing to this !
> On 16 September 2016 at 02:40, Ariel Rokem <arokem at gmail.com> wrote:
> On Thu, Sep 15, 2016 at 1:33 PM, Kirstie Whitaker <kw401 at cam.ac.uk>
> wrote:
> Hi John,
> I'm travelling at the moment but I've had problems with pysurfer
> showing beautiful brains on the screen but only saving a black box
> to file. It happened right after our systems admin updated a few
> things but I haven't been able to get a clear list from him of what
> changed except: everything should work.
> My point with this email is please do share back what you
> learn.....even if it ends up being not a pysurfer problem. At the
> moment my workaround is to move everything I do to a different
> cluster that works!! Non efficient to say the least!
> Thank you
> Kirstie
> Sent from my iPhone, please excuse any typos or excessive brevity
> On 15 Sep 2016, at 12:44, John Pellman <pellman.john at gmail.com>
> wrote:
> I've had at this a little bit more and my current suspicion is
> that this behavior is the result of an interaction between our
> remote desktop service (x2go) and Mayavi.
> I created a an identical Miniconda environment for Pysurfer on
> both our server and my laptop and ran the following code to
> test this theory:
> # The Basic Visualization demo from the Pysurfer gallery.
> from surfer import Brain
> print(__doc__)
> """
> Define the three important variables.
> Note that these are the first three positional arguments
> in tksurfer (and pysurfer for that matter).
> """
> subject_id = 'fsaverage'
> hemi = 'lh'
> surface = 'inflated'
> """
> Call the Brain object constructor with these
> parameters to initialize the visualization session.
> """
> brain = Brain(subject_id, hemi, surface)
> # Save an image out to /tmp
> print 'Saving out an image to /tmp using Brain.save_image.'
> brain.save_image('/tmp/brain.png')
> # Looking at just the screenshot method of pysurfer's Brain
> object.
> # This is called by save_image and is fed into
> scipy.misc.imsave.
> # If the boolean expression evaluated here is true, then
> only a black
> # background is being fed into scipy's misc.imsave method
> for evaluation.
> x = brain.screenshot()
> print 'Test pysurfer\'s Brain.screenshot.'
> if sum(x.flatten()==0)!=len(x.flatten()):
> print 'Pass'
> else:
> print 'Fail'
> # Looking at the Mayavi mlab.screenshot method.
> # This is called by screenshot_single, which is called by
> Brain's screenshot.
> # If the boolean expression evaluated here is true, then
> only a black
> # background is being fed into Brain.screenshot()
> from mayavi import mlab
> x = mlab.screenshot(brain.brain_matrix[0,0]._f, 'rgb',
> False)
> print 'Test mayavi\'s mlab.screenshot'
> if sum(x.flatten()==0)!=len(x.flatten()):
> print 'Pass'
> else:
> print 'Fail'
> On the server through an x2go session both Brain.screenshot and
> mlab.screenshot failed to produce a non-blank image, while on
> my laptop's local environment both of these methods did produce
> the desired output (i.e., there were some nonzero outputs).
> Since this doesn't seem to be an error with pysurfer in
> particular, I'm going to proceed to see if anyone using Mayavi
> with x2go or nx has encountered similar issues by querying
> their forums / issue pages. I just wanted to leave this here
> if someone else encounters the same issue in the future.
> A shot in the dark: Could it be something to do with running headless?
> Maybe running this under XVFB (e.g. through xvfbwrapper) would help?
> Ariel
>
> --John
> On Tue, Sep 13, 2016 at 1:24 PM, John Pellman <
> pellman.john at gmail.com> wrote:
> It looks like it might be related to the following issue
> described at StackOverflow:
> http://stackoverflow.com/questions/16543634/
> mayavi-mlab-savefig-gives-an-empty-image
> On Mon, Sep 12, 2016 at 2:00 PM, John Pellman <
> pellman.john at gmail.com> wrote:
> Hi all,
> I'm encountering a peculiar Pysurfer error on our
> server and I was wondering if anyone has encountered
> anything similar or might have some insight into how I
> can tackle it. Basically, when our researchers try to
> save a png image using Brain.save_image() or
> Brain.save_imageset() the images produced only contain
> the background color (as you may have inferred from the
> subject line). I've traced this back to Scipy method
> (scipy.misc.imsave), but it looks like this would only
> output an empty png if the image passed in were
> completely zeroed out. Our setup uses the following
> versions of pysurfer/its dependencies:
> Numpy: 1.10.0.dev0+1fe98ff
> Scipy: 0.17.0.dev0+f2f6e48
> Ipython: 3.1.0
> nibabel: 2.0.0
> Mayavi: 4.4.2
> matplotlib: 1.4.3
> PIL: 1.1.7
> Pysurfer: 0.5
> This setup is running within a Miniconda environment
> using Python 2.7.11. I'm uncertain if this is related,
> but running the example code here produces the
> following warning:
> (ipython:20765): Gdk-WARNING **: /build/buildd/
> gtk+2.0-2.24.27/gdk/x11/gdkdrawable-x11.c:952 drawable
> is not a pixmap or window
> Any insight would be greatly appreciated.
> Best,
> John Pellman
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
--
Gael Varoquaux
Researcher, INRIA Parietal
NeuroSpin/CEA Saclay , Bat 145, 91191 Gif-sur-Yvette France
Phone: ++ 33-1-69-08-79-68
http://gael-varoquaux.info http://twitter.com/GaelVaroquaux
More information about the Neuroimaging
mailing list