Impact of Parameter Disentanglement on Collaborative Alignment

Tianyu Song, Alejandro Martin-Gomez, Qiaochu Wang, Arian Mehrfard, Javad Fotouhi, Daniel Roth, Ulrich Eck, Nassir Navab

Abstract: The interactive alignment of real and virtual content in AR is often non-trivial. Positional errors along the users view direction frequently lead to the misjudgment of the objects depth. This work takes advantage of alternative users' viewpoints in collaborative settings to mitigate these errors. Furthermore, we systematically restrict the parameters used to control the virtual contents pose and investigate the impact of sharing and disentangling such parameters. Results from this work show that alignment schemes that disentangle the control parameters improve overall alignment accuracy with a similar workload for the users and no significant increase in execution time.