Evaluation of Segmentation Algorithms for Embodiment Improvement in an XR Application

In extended reality applications there is an increasing interest in combining virtual environments with realistic avatars and objects so the user can easily interact with them. Segmentation algorithms are already used to include video-based self-avatars. We propose using these algorithms to combine the real body of the user and surrounding objects to improve the sense of embodiment and the ability to interact in the virtual environment in a table etiquette use case. Moreover, we present the results of objective and subjective evaluations of the implemented algorithms to study the correlation between these metrics. We also highlight that the subjective evaluation includes participants with intellectual disabilities as potential users of the technology. This is crucial for ensuring that development and implementations account for their experiences, aiming to create more inclusive applications and evaluation methodologies.