Eye-Tracking over Source Code : A Benchmarking Extension to Evaluate the Accuracy ofEyesOnTheCode

Abstract: Context: This report covers the development of a benchmarking extension to the eye-tracking software ”EyesOnTheCode” that utilizes WebGazer. The extension makes it possible tobenchmark the software and for researchers to quantify externalfactors’ impact on eye-tracking.Objective: To create a benchmarking extension to EyesOnTheCode that measures the accuracy of its readings. Additionally,to create a data analysis tool that assists researchers in interpretingthe data more easily.Approach: The benchmarking software displays a movingobject on the screen that the user follows with their gaze. Thesoftware records data throughout the benchmarking sessionthat can then be analyzed through the data analysis tool. Thetool makes it possible to generate charts that can be used forresearch purposes.Results: The use of reading glasses increased the Euclideandistance by 44.55 pixels, with a decrease in accuracy fromthe control test by 20.5 percent. A decrease in web cameraresolution from 1920 by 1080 to 1280 by 720 pixels reducedthe accuracy by an average of 77.5 pixels or roughly 35.7percent. Increasing the web camera resolution from 1920 by1080 to 3840 by 2160 pixels decreased the Euclidean distanceby 27.66 pixels, increasing accuracy by roughly 12.7 percent.Conclusion: Data polled from the software indicates thatusing reading glasses and low web camera resolution negativelyimpacts eye-tracking. Further work and more test samples must bemade to verify the data, but the benchmarking software showspromising results. More features may need to be implemented tomake the software more usable for researchers.