no description available

Xiaoyi Yang

Xueling Zhang

Rochester Institute of Technology

Augmented Reality (AR) software development frameworks typically provide virtual reality scenes for testing purposes. To make sure AR apps meet users’ expectation, precision of virtual object placement is an important measurement of AR applications’ usability during AR software testing. Within virtual reality scenes, although a test script can automatically move the camera to view different parts of the scene and to place virtual objects at different locations, the testing process is still often manual because a human tester needs to either watch the test execution or watch videos / screenshots recorded during test execution to decide whether an object placement is noticeably imprecise. In this paper, we develop a novel technique, PredART, to explore whether it is possible to predict whether the placement of an object is noticeably imprecise by human users. The prediction results can be used for automatic assertions in AR software testing and raise warnings to human testers only when a potential imprecise placement is found. Our evaluation shows that PredART is able to predict noticeable placement errors with a F-score of 75%.

Tahmid Rafi

University of Texas at San Antonio

United States

Xueling Zhang

Rochester Institute of Technology

Xiaoyin Wang

University of Texas at San Antonio