[solved] Unable to set backdrop in Eyelink 1000 using OS portable 0.27.3

Comments

• Daniel

  May 2014 edited 4:52PM

  Hi gerardo,

  You need to call the set_backdrop() function before you have called eyelink.start_recording() or before the eyelink_start_recording item is executed. Somehow the backdrop is not shown if you call the set_backdrop() function during recording of eye movements. Also, this method is very slow, so make sure using backdrop functionalities does not impact your response measurements.

  
• gerardo

  June 2014 edited June 2014

  Hi, Daniel, thank you for your fast answer. I've tried doing as you said, moving the set_backdrop() line before the eyelink_start_recording item in the trial sequence, but still, the stimuli are not displayed in the host pc. Everything seems to work fine, and the program gives no exception; but the canvas is not converted and/or sent to the host pc. I fear I might be doing something wrong, but I don't know what. My trial sequence is:

  eyelink_drift_correct set_canv (inline script with just the line I mentioned in my first post)

  eyelink_start_recording

  log_stim (a message to the log)

  image_pair (a canvas consisting in two random bmp images over a grey background)

  eyelink_stop_recording

  blank (a blank canvas)

  prepare_canv (inline script that copies image_pair canvas and prepares it to be set)

  logger

  I've tried writing all the backdrop code in the run phase before eyelink_start_recording, which doesn't result in any visible change; and to directly copy and set the backdrop without using prepare_backdrop(). The documentation seems to suggest that this is possible, but I get an bad argument exception. In my case, it seems that the set_backdrop() doesn't accept an unprepared canvas.

  So, this is my situation. Do you have any idea what might be happening now? Could you give me an example of an experiment using the backdrop functionalities, so I can check if I've done everything right?

  Thanks for your help!
• Daniel

  June 2014 edited 4:52PM

  Hi Gerardo,

  last time I did it, this experiment structure worked:

  

  in stimulus_display, I create a canvas in the prepare phase and when I have finished drawing I prepare it as a backdrop image with the following line at the bottom of the script:

  exp.bd = exp.eyelink.prepare_backdrop(canvas)
  

  Then as the first thing in the run phase, I send the backdrop to the Eyelink. In set_backdrop in the run (!) phase I call:

  exp.eyelink.set_backdrop(exp.bd)
  

  I don't think it matters if you do this before or after drift_correct (I did it before in this example) as long as you do it before you start recording. I don't know why it doesn't work for your eyelink installation. Is it a very old one? What type of Eyelink is it?

  
• gerardo

  June 2014 edited 4:52PM

  Thanks again. I will try in a few hours to see if I manage to make it work. I'm working with a desktop mount of an Eyelink 1000. I was not here when they bought it and installed, so I really don't know more about it.

  The only difference I can see from your experiment is that you create your own canvas while I use a sketchpad item that then I copy with self.copy_sketchpad(). I might try creating a simple canvas just to see if that has anything to do with it.
• gerardo

  June 2014 edited 4:52PM

  No luck
  My experiment was set exactly as yours, even with the

  exp.eyelink.set_backdrop()

  instruction in the run phase before the drift correction, at the beginning of the trial.

  I've also tried directly coding a new canvas, with no luck. I've tried one of the example experiments provided by eyelink with their own software, and the backdrop is properly displayed; so it must be something on the OS end. Could you by any chance send me your working experiment (or part of it), so I can see if the problem remains?

  I'm using an eyelink 1000 with the 4.51 version of the host pc software. I believe it was purchased 2-4 years ago. No one has ever used it before (which is a shame), so I really don't know if they did something wrong when installing it; but everything else works fine. Gaze is recorded, stimuli are displayed in the display pc, edf files are created and stored in both pcs...

  I'm using OS portable 0.27.3. I've finally solved my silly problem with OS 2.8.1; but still, every time a test is aborted, 2.8.1 kind of freezes (you can close OS and even save the experiment when prompted, but everything is greyed out and the GUI doesn't respond to any input). On the other hand, 0.27.3 requires 2 fixations to proceed when in drift_correct, I don't know why. 2.8.1 or the examples provided by eyelink don't show this behaviour. In neither of these two versions does setting the backdrop work. Do you think that running from source might solve the problem?

  I really don't know what else to do. I went with OS because I wanted to save time and keep my colleagues' pc clean, without unneeded installations. I'm still happy with the choice, but it is frustrating not to be able to set the backdrop. In any case, as it is not a vital part of my experiment, I might end up just forgetting about it, but it would be nice if we could figure out what's wrong.

  Anyway, thanks for your time.
• Daniel

  June 2014 edited 4:52PM

  I also had such problems here. We have several Eyelink installations, some of which are also of different models and makes. The backdrop functionality is a really 'iffy' feature of the Eyelink software, which is also badly implemented (sloooww!). Because of these differences in behaviour per system, it is really hard to pinpoint any problems.

  The last thing I can do, is send you an experiment of mine, of which I know the backdrop functionality worked for me. If it doesn't work for you, you know it is your setup. If it does work, you must be doing something slightly different in your experiment which breaks the backdrop stuff altogether.

  Regarding the problem you describe in your middle paragraph: this seems like a different or new problem altogether, and maybe it is a good idea to open a separate forum topic for this.

  I also noticed that you described that you are not running OpenSesame from source. I thought the whole eyelink plugin only worked when you ran OpenSesame from source and was not available for the packaged OS versions. Due to licensing restrictions, it is not allowed to package pylink (the python eyelink module) with OpenSesame, so you should not be able to work with the eyelink at all with the packaged version. Regardless of this, it is worth a try to use a portable version of OpenSesame, which basically is the same thing as running from source.

  
• gerardo

  June 2014 edited 4:52PM

  Well, I have to apologize for my misuse of computing jargon. I'm indeed running the portable versions of OS; and, therefore, kind of running from source already. So, checking your experiment seems to be the best (and only?) option. I've made my e-mail visible in my profile, so you can send me the file to that address.

  At the moment, I want to focus on this issue (and not to overload the forum with silly questions ). I'll be looking at these other problems I mentioned and, if they persist, I will create a new topic.
• Daniel

  June 2014 edited 4:52PM

  Hi Gerardo, I've been trying to find an appropriate experiment, but most I have readily available involve large image sets which I can't easily send along. I will try to create a simpler experiment that uses backdrops and see if it works here before I send it to you.

  
• gerardo

  June 2014 edited 4:52PM

  Thank you, Daniel, I'll wait for it.

  In the meanwhile, I've decided to move on and do the pilot testing without the backdrop functionality. Everything else has worked fine, except for a few quirks that I'll write about in a different post this afternoon, if I have the time.
• gerardo

  June 2014 edited June 2014

  Hi, Daniel,
  I was thinking about it and I realized that, when I first run OpenSesame 0.27.3, I had to remove line 898 from libeyelink.py, as it was negating the if statement and causing the program to crash:

  if hasattr(el,"bitmap2DBackdrop"):
      send_backdrop = el.bitmap2DBackdrop
  else:
      send_backdrop = el.bitmapBackdrop
  send_backdrop = el.bitmap2DBackdrop
  

  I think this might be related to my problems with set_backdrop. Could it be that el.bitmapBackdrop, or the lack of el.bitmap2DBackdrop, is causing it?
• Daniel

  June 2014 edited 4:52PM

  Hi Gerardo,

  of course, you're right. I forgot all about a conversation a while ago with the SR helpdesk during which they built a custom pylink module for me. It is weird that they haven't implemented these changes of my custom module in the main version of pylink, but I guess their maintenance is quite lacking anyway as there hasn't been an update to the main pylink version for a long while now.

  If I get back to work on Monday I will send you my pylink module, which will hopefully make things work for you.

  
• ambraferrari

  July 2014 edited 4:52PM

  Hi Gerardo and Daniel,

  I'm facing the same problem with 2.8.1 version.
  I'm trying to connect an Eyelink 1000 with a OS experiment which consists of a unique (big) inline_script component.
  When I put the PyGaze components (calibration, start recording, log, stop recording) in the Experiment sequence without the inline_script component, it gives me the error: "The EyeLink cannot handle filenames longer than eight characters". How did you solve it?
  Then, when I also put the inline_script component (after calibration), OS simply ignores the calibration component, do not show the error above and starts my experiment.
  Can you help me?

  Thanks, Ambra
• sebastiaan

  July 2014 edited 4:52PM

  Hi Ambra,

  When I put the PyGaze components (calibration, start recording, log, stop recording) in the Experiment sequence without the inline_script component, it gives me the error: "The EyeLink cannot handle filenames longer than eight characters". How did you solve it?

  The error message basically says it all: The filename (without the .csv extension) should be eight characters or less. So subject-0.csv is ok (8 characters), but subject-10.csv is not (9 characters).

  Cheers,
  Sebastiaan

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!
  
• ambraferrari

  July 2014 edited July 2014

  Hi Sebastiaan,

  thanks for your reply, we fixed the error "The EyeLink cannot handle filenames longer than eight characters".
  We are now able to open the connection with the Eyelink. In the Experiment there are few components in this order:

  1. pygaze_init
  2. pygaze_start_recording
  3. inline_script (with my whole experiment in python language)
  4. pygaze_log
  5. pygaze_stop_recording.

  Calibration and validation work fine, then the program asks for 'q' press to go back to menu. We press it once, then the stimulation PC display turns black and we don't see our experiment running. We are using PsychoPy back end. We also tried the example called pygaze_plugins provided in WinPython-OpenSesame-2.8.1\OpenSesame\opensesame-2.8.1\examples, it uses legacy back end and it works properly.

  Do you have any suggestions about this issue?

  Cheers, Stefania and Ambra
• Edwin

  July 2014 edited 4:52PM

  Have you tried pressing 'q' a second time? The blank screen you are seeing might well be the menu, which is not drawn due to some underlying error. To confirm this theory, either press 'q' a second time, or press 'Enter' after pressing 'q' the first time (this should show you the eye image; pressing 'Enter' again should make you return to the visible menu).
• ambraferrari

  July 2014 edited 4:52PM

  Hi Edwin,
  thank for your reply, we finally succeeded in running the experiment! We have to press the 'q' a second time to start the inline_script that follows calibration. Then everything works properly.
  Cheers, Ambra
• hanzh

  October 2017 edited October 2017

  Hi all,

  I don't know if anyone is still following this, but I also have the problem of unable to set the backdrop. I inserted an inline_script based on functions from the old Pygaze and it worked for both pygame and expyriment backend (didn't try on psychopy, tho)

  The code basically reads an image file from the filepool and send it to the host PC. So before the experiment loop starts:

  import pylink
  from PIL import Image
  import numpy as np
  
  def prepare_backdrop(im):
      # open image
      # for some weird reason the image is not displayed in its original color
      # perhaps a transparency issue (images of texts were exported from PowerPoint)
      img = Image.open(im).convert('I') 
      # convert image
      img_arr = np.array(img).tolist()
  
      return (img_arr,exp.width,exp.height)
  
  
  def set_backdrop(backdrop):
      starttime = exp.time()
  
      # call prepare_backdrop to convert image
      backdrop = prepare_backdrop(backdrop)
  
      # check if format is correct
      if type(backdrop)==tuple:
          if len(backdrop) != 3 or type (backdrop[0]) != list or \
          type(backdrop[1]) != int or type (backdrop[2]) != int:
              print 'hahahah', type (backdrop[0]),type(backdrop[1]),type(backdrop[2])
              raise Exception('Invalid tuple format')
          else:
              el = pylink.getEYELINK()
              # the other option "el.bitmap2DBackdrop" do not work for me
              send_backdrop = el.bitmapBackdrop
              img = backdrop[0]
              width = backdrop[1]
              height = backdrop[2]
              send_backdrop(width,height,img,0,0,width,height,0,0,pylink.BX_MAXCONTRAST)
      else:
          raise Exception(u'Unable to send backdrop')
  
      return exp.time() - starttime
  

  Then, within the experiment loop (in the prepare tab, before drift correction and start_recording) do the following:

  path = exp.pool[var.pic]
  lag = set_backdrop(path)
  

  I did notice a non-trivial lag. For Pygame it was about >200ms for expyriment it was about 100~200ms. However, because the image was sent to the host PC during the prepare phase in each trial it shouldn't affect too much (@sebastiaan is this correct?). Before the stimulus shows up, I also have a buffer blank image for 250ms after "start_recording", to equalize potential lags resulted from these communications. For most of the time, the image would first show up on the host PC.

  Hope this helps,
  Han
• sebastiaan

  October 2017

  However, because the image was sent to the host PC during the prepare phase in each trial it shouldn't affect too much (@sebastiaan is this correct?)

  That's correct—in that case it only causes a lag between trials.

  And thanks for sharing your code!

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!