Running the validation procedure

Following the calibration procedure you can perform a validation. Running a validation is recommended because firstly, it informs you of the quality of the current calibration, and secondly, since a calibration represents a set of gaze mapping functions, we can select the best one, given the current validation data.

Click the Validate button or press v or V to start. To run a validation without selecting the best gaze mapping function post validation, press Ctrl+v.

On the test screen, targets will be presented one after another as defined in the settings. To the participant the procedure is the same as for calibration.

On succesful completion of the validation two info signs will appear below the Validate button. Hovering the mouse cursor over these info signs will show the validation results.

Hovering the mouse cursor over the first info sign shows a qualitative indication of how many validation points are good, fair, poor, bad or invalid and it also gives the average validation errors. Ideally, you want the average error to lie below 0.5°.

Hovering the mouse cursor over the second info sign displays the validation errors for each point graphically (see example output). For each validation point there will be a white circle with a radius of 0.5° and ellipses showing the gaze estimation for the right, left and mean eye. Ideally the centers of the estimated gaze points should lie within the white circle.

Repeat the validation if there one or more points with a large error (say > 2° or perhaps 1°) since that would be a sign the validation partially went wrong. If there are large errors and revalidation doesn’t help, then likely a recalibration is necessary. Note, repeating the procedure over and over until all points are good is senseless.

See Validate in the Eyevec-control user manual for example output.