Report ID
Report Authors
Adam Ibrahim, John B. Lanier, Brandon Huynh, John O'Donovan, Tobias Höllerer
Report Date

We investigate the possibility of running object recognition tasks on the Microsoft HoloLens Augmented Reality (AR) headset in real time. Two different approaches are implemented: one using the Microsoft Project Oxford API, the other using a pre-trained neural network (YOLO) running on a local server. We find that while there are workarounds for the latency, the main obstacle to real-time performance is the HoloLens itself.

tr2017-04.pdf990.28 KB