readmeAdvancedCalibration_auto.m

% this demo code is part of Titta, a toolbox providing convenient access to
% eye tracking functionality using Tobii eye trackers
%
% Titta can be found at https://github.com/dcnieho/Titta. Check there for
% the latest version.
% When using Titta, please cite the following paper:
%
% Niehorster, D.C., Andersson, R. & Nystrom, M., (2020). Titta: A toolbox
% for creating Psychtoolbox and Psychopy experiments with Tobii eye
% trackers. Behavior Research Methods.
% doi: https://doi.org/10.3758/s13428-020-01358-8
%
% This example uses the advanced calibration interface
% (Titta.calibrateAdvanced). When using this interface, please also cite
% the following paper:
%
% Niehorster, D.C., Whitham, W., Lake, B.R., Schapiro, S.J., Andolina, I.M.
% & Yorzinski, J.L. (2024). Enhancing eye tracking for nonhuman primates
% and other subjects unable to follow instructions: Adaptive calibration
% and validation of Tobii eye trackers with the Titta toolbox. Behavior
% Research Methods. doi: https://doi.org/10.3758/s13428-024-02540-y

% This version of readme.m demonstrates operation with separate
% presentation and operator screens. It furthermore demonstrates Titta's
% advanced calibration mode that is designed for working with subjects who
% are unable to follow instructions, such as nonhuman primates and infants.
% This version uses a controller for automatically calibrating, in multiple
% steps (first a center calibration points is shown, data collected for it
% and calibration appied based on it, then in two more steps, data is
% collected for two times three other points).
% A dummy class providing rewards when looking at the calibration points or
% inside a red circle shown after calibration is also implemented for demo
% purposes.
% 
% NB: some care is taken to not update operator screen during timing
% critical bits of main script
% NB: this code assumes main and secondary screen have the same resolution.
% Titta's setup displays work fine if this is not the case, but the
% real-time gaze display during the mock experiment is not built for that.
% So if your two monitors have different resolutions, either adjust the
% code, or look into solutions e.g. with PsychImaging()'s 'UsePanelFitter'.

clear all
sca

DEBUGlevel              = 0;
fixClrs                 = [0 255];
bgClr                   = 127;
eyeColors               = {[255 127 0],[0 95 191]}; % for live data view on operator screen
scrParticipant          = 1;
scrOperator             = 2;
useWindowedOperatorScreen = false;  % if true, a windowed operator display smaller than the whole screen is made
% task parameters
fixTime                 = .5;
imageTime               = 4;
% live view parameters
dataWindowDur           = .5;   % s

% You can run addTittaToPath once to "install" it, or you can simply add a
% call to it in your script so each time you want to use Titta, it is
% ensured it is on path
home = cd;
cd ..;
addTittaToPath;
cd(home);

try
    eyeColors = cellfun(@color2RGBA,eyeColors,'uni',false);
    
    % get setup struct (can edit that of course):
    settings = Titta.getDefaults('Tobii Pro Spectrum');
    % request some debug output to command window, can skip for normal use
    settings.debugMode      = true;
    % customize colors of setup and calibration interface (yes, colors of
    % everything can be set, so there is a lot here).
    % operator screen
    settings.UI.advcal.bgColor              = bgClr;
    settings.UI.advcal.fixBackColor         = fixClrs(1);
    settings.UI.advcal.fixFrontColor        = fixClrs(2);
    settings.UI.advcal.fixPoint.text.color  = fixClrs(1);
    settings.UI.advcal.avg.text.color       = fixClrs(1);
    settings.UI.advcal.instruct.color       = fixClrs(1);
    settings.UI.advcal.gazeHistoryDuration  = dataWindowDur;
    settings.UI.advcal.fixPoint.text.size   = 24;
    % calibration display: add two more points so we have a square grid
    % with center point
    settings.advcal.cal.pointPos = [settings.advcal.cal.pointPos; .5, .1; .5, .9];
    % calibration display: custom calibration drawer
    calViz                      = MultiTargetCalibrationDisplay();
    settings.advcal.drawFunction= @calViz.doDraw;
    calViz.bgColor              = bgClr;
    calViz.fixBackColor         = fixClrs(1);
    calViz.fixFrontColor        = fixClrs(2);
    % calibration logic: custom controller
    rewardProvider = DemoRewardProvider();
    rewardProvider.dutyCycle = 170; % ms
    calController = MultiStepCalController([],calViz,[],rewardProvider);
    % hook up our controller with the state notifications provided by
    % Titta.calibrateAdvanced (request extended notification) so that the
    % calibration controller can keep track of whats going on and issue
    % appropriate commands.
    settings.advcal.cal.pointNotifyFunction = @calController.receiveUpdate;
    settings.advcal.val.pointNotifyFunction = @calController.receiveUpdate;
    settings.advcal.cal.useExtendedNotify = true;
    settings.advcal.val.useExtendedNotify = true;
    % show the button to start the controller.
    settings.UI.button.advcal.toggAuto.visible = true;
    calPoints = {3,[2 5 6],[1 4 7]};    % show calibration points in three steps. Denote which points in which steps
    calPoss   = cellfun(@(x) settings.advcal.cal.pointPos(x,:),calPoints,'uni',false);
    calController.setCalPoints(calPoints,calPoss);
    calController.calAfterEachStep = true;  % tell controller to update calibration after each step is completed
    if DEBUGlevel>0
        calController.logTypes  = 1+2*(DEBUGlevel==2)+4;    % always log actions calController is taking and reward state changes. Additionally log info about received commands when DEBUGlevel==2
    end
    calController.logReceiver = 1;                          % 1: log to Titta messages

    
    % init
    EThndl          = Titta(settings);
    EThndl.init();
    calController.EThndl = EThndl;
    nLiveDataPoint  = ceil(dataWindowDur*EThndl.frequency);
    
    if DEBUGlevel>1
        % make screen partially transparent on OSX and windows vista or
        % higher, so we can debug.
        PsychDebugWindowConfiguration;
    end
    if DEBUGlevel
        % Be pretty verbose about information and hints to optimize your code and system.
        Screen('Preference', 'Verbosity', 4);
    else
        % Only output critical errors and warnings.
        Screen('Preference', 'Verbosity', 2);
    end
    Screen('Preference', 'SyncTestSettings', 0.002);    % the systems are a little noisy, give the test a little more leeway
    Priority(1);
    [wpntP,winRectP] = PsychImaging('OpenWindow', scrParticipant, bgClr, [], [], [], [], 4);
    if useWindowedOperatorScreen
        wrect  = Screen('GlobalRect', scrOperator);
        [w, h] = Screen('WindowSize', scrOperator);
        wrect  = CenterRect([w*.1 h*.1 w*.9 h*.9],wrect);
        [wpntO,winRectO] = PsychImaging('OpenWindow', scrOperator, bgClr, wrect, [], [], [], 4, [], kPsychGUIWindow);
    else
        [wpntO,winRectO] = PsychImaging('OpenWindow', scrOperator, bgClr, [], [], [], [], 4);
    end
    hz=Screen('NominalFrameRate', wpntP);
    Screen('BlendFunction', wpntP, GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
    Screen('BlendFunction', wpntO, GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
    Screen('Preference', 'TextAlphaBlending', 1);
    Screen('Preference', 'TextAntiAliasing', 2);
    % This preference setting selects the high quality text renderer on
    % each operating system: It is not really needed, as the high quality
    % renderer is the default on all operating systems, so this is more of
    % a "better safe than sorry" setting.
    Screen('Preference', 'TextRenderer', 1);
    KbName('UnifyKeyNames');    % for correct operation of the setup/calibration interface, calling this is required

    calController.scrRes = winRectP(3:4);

    
    % do calibration
    try
        ListenChar(-1);
    catch ME
        % old PTBs don't have mode -1, use 2 instead which also supresses
        % keypresses from leaking through to matlab
        ListenChar(2);
    end
    tobii.calVal{1} = EThndl.calibrateAdvanced([wpntP wpntO],[],calController);
    ListenChar(0);
    
    % prep stimuli (get rabbits) - preload these before the trials to
    % ensure good timing
    rabbits = loadStimuliFromFolder(fullfile(PsychtoolboxRoot,'PsychDemos'),{'konijntjes1024x768.jpg','konijntjes1024x768blur.jpg'},wpntP,winRectP(3:4));
    
    
    % later:
    EThndl.buffer.start('gaze');
    WaitSecs(.8);   % wait for eye tracker to start and gaze to be picked up
     
    % send message into ET data file
    EThndl.sendMessage('test');
    
    % First draw a fixation point
    Screen('gluDisk',wpntP,fixClrs(1),winRectP(3)/2,winRectP(4)/2,round(winRectP(3)/100));
    startT = Screen('Flip',wpntP);
    % log when fixation dot appeared in eye-tracker time. NB:
    % system_timestamp of the Tobii data uses the same clock as
    % PsychToolbox, so startT as returned by Screen('Flip') can be used
    % directly to segment eye tracking data
    EThndl.sendMessage('FIX ON',startT);
    nextFlipT   = startT+fixTime-1/hz/2;
    
    % now update also operator screen, once timing critical bit is done
    % if we still have enough time till next flipT, update operator display
    while nextFlipT-GetSecs()>2/hz   % arbitrarily decide two frames is enough headway
        Screen('gluDisk',wpntO,fixClrs(1),winRectO(3)/2,winRectO(4)/2,round(winRectO(3)/100));
        drawLiveData(wpntO,EThndl.buffer.peekN('gaze',nLiveDataPoint),dataWindowDur,eyeColors{:},4,winRectO(3:4));
        Screen('Flip',wpntO);
    end
        
    
    % show on screen and log when it was shown in eye-tracker time.
    % NB: by setting a deadline for the flip, we ensure that the previous
    % screen (fixation point) stays visible for the indicated amount of
    % time. See PsychToolbox demos for further elaboration on this way of
    % timing your script.
    Screen('DrawTexture',wpntP,rabbits(1).tex,[],rabbits(1).scrRect);
    imgT = Screen('Flip',wpntP,nextFlipT);
    EThndl.sendMessage(sprintf('STIM ON: %s [%.0f %.0f %.0f %.0f]',rabbits(1).fInfo.name,rabbits(1).scrRect),imgT);
    nextFlipT = imgT+imageTime-1/hz/2;
    
    % now update also operator screen, once timing critical bit is done
    % if we still have enough time till next flipT, update operator display
    while nextFlipT-GetSecs()>2/hz   % arbitrarily decide two frames is enough headway
        Screen('DrawTexture',wpntO,rabbits(1).tex);
        drawLiveData(wpntO,EThndl.buffer.peekN('gaze',nLiveDataPoint),dataWindowDur,eyeColors{:},4,winRectO(3:4));
        Screen('Flip',wpntO);
    end
    
    % record x seconds of data, then clear screen. Indicate stimulus
    % removed, clean up
    endT = Screen('Flip',wpntP,nextFlipT);
    EThndl.sendMessage(sprintf('STIM OFF: %s',rabbits(1).fInfo.name),endT);
    Screen('Close',rabbits(1).tex);
    nextFlipT = endT+1; % less precise, about 1s give or take a frame, is fine
    
    % now update also operator screen, once timing critical bit is done
    % if we still have enough time till next flipT, update operator display
    while nextFlipT-GetSecs()>2/hz   % arbitrarily decide two frames is enough headway
        drawLiveData(wpntO,EThndl.buffer.peekN('gaze',nLiveDataPoint),dataWindowDur,eyeColors{:},4,winRectO(3:4));
        Screen('Flip',wpntO);
    end
    
    % repeat the above but show a different image. lets also record some
    % eye images, if supported on connected eye tracker
    if EThndl.buffer.hasStream('eyeImage')
       EThndl.buffer.start('eyeImage');
    end
    % 1. fixation point
    Screen('gluDisk',wpntP,fixClrs(1),winRectP(3)/2,winRectP(4)/2,round(winRectP(3)/100));
    startT      = Screen('Flip',wpntP,nextFlipT);
    EThndl.sendMessage('FIX ON',startT);
    nextFlipT   = startT+fixTime-1/hz/2;
    while nextFlipT-GetSecs()>2/hz   % arbitrarily decide two frames is enough headway
        Screen('gluDisk',wpntO,fixClrs(1),winRectO(3)/2,winRectO(4)/2,round(winRectO(3)/100));
        drawLiveData(wpntO,EThndl.buffer.peekN('gaze',nLiveDataPoint),dataWindowDur,eyeColors{:},4,winRectO(3:4));
        Screen('Flip',wpntO);
    end
    % 2. image
    Screen('DrawTexture',wpntP,rabbits(2).tex,[],rabbits(2).scrRect);
    imgT = Screen('Flip',wpntP,startT+fixTime-1/hz/2);                  % bit of slack to make sure requested presentation time can be achieved
    EThndl.sendMessage(sprintf('STIM ON: %s [%.0f %.0f %.0f %.0f]',rabbits(2).fInfo.name,rabbits(2).scrRect),imgT);
    nextFlipT = imgT+imageTime-1/hz/2;
    while nextFlipT-GetSecs()>2/hz   % arbitrarily decide two frames is enough headway
        Screen('DrawTexture',wpntO,rabbits(2).tex);
        drawLiveData(wpntO,EThndl.buffer.peekN('gaze',nLiveDataPoint),dataWindowDur,eyeColors{:},4,winRectO(3:4));
        Screen('Flip',wpntO);
    end
    
    % 3. end recording after x seconds of data again, clear screen.
    endT = Screen('Flip',wpntP,nextFlipT);
    EThndl.sendMessage(sprintf('STIM OFF: %s',rabbits(2).fInfo.name),endT);
    Screen('Close',rabbits(2).tex);
    Screen('Flip',wpntO);
    
    % stop recording
    if EThndl.buffer.hasStream('eyeImage')
        EThndl.buffer.stop('eyeImage');
    end
    EThndl.buffer.stop('gaze');
    
    % save data to mat file, adding info about the experiment
    dat = EThndl.collectSessionData();
    dat.expt.resolution = winRectP(3:4);
    dat.expt.stim       = rabbits;
    EThndl.saveData(dat, fullfile(cd,'t'), true);
    % if you want to (also) save the data to Apache Parquet and json files
    % that can easily be read in Python (Apache Parquet files are supported
    % by Pandas), use:
    % EThndl.saveDataToParquet(dat, fullfile(cd,'t'), true);
    % All gaze data columns and messages can be dumped to tsv files using:
    % EThndl.saveGazeDataToTSV(dat, fullfile(cd,'t'), true);
    
    % shut down
    EThndl.deInit();
catch me
    sca
    ListenChar(0);
    rethrow(me)
end
sca