Recording of repeatable AR sessions exemplified by ARKit and Unity
Tobias Schwandt, Christian Kunert, Wolfgang Broll · 2024
The development and testing of Augmented Reality (AR) applications often involve non-reproducible conditions, making it challenging to conduct consistent and repeatable experiments. To address this issue, we developed ARC (AR Capture), a native Swift application utilizing ARKit to record AR sessions, and ARW (AR Workstation), a Unity-based tool for replaying these sessions. ARC captures RGB-D data, tracking information, and lighting properties, storing them in a machine-and human-readable format. In ARW, these recorded sessions can be loaded, played back, and manipulated within Unity, allowing developers to register for events, integrate custom algorithms, and conduct reproducible analyses. Our approach enables the replay of AR scenes multiple times under varying conditions, providing a robust platform for studies requiring reproducible scenarios. This capability facilitates rigorous testing of new algorithms in areas such as geometric and photo-metric reconstruction. Moreover, the integration with Unity allows for sophisticated experimentation on desktop computers, including complex voxel structure generation, ray tracing, and neural network applications. This work highlights the potential for ARC and ARW to enhance research in AR by providing a reliable framework for the reproducible testing of AR applications. Future updates will focus on expanding the range of captured data and improving playback fidelity, further broadening the scope of possible AR research and development.