Inside-out tracking for VR/XR

The Spectacular AI SDK can provide accurate and low-latency 6-DoF pose tracking for AR, VR and XR headsets, without external lighthouses, the solution is computationally lightweight and can run on an embedded ARM processor in the headset.

For low-cost and experimental setups, the OAK-D devices can be useful for prototyping this use case, see, e.g., the mixed reality case in the SDK examples repository.

In general, we recommend a system with at least two hardware-synchronized fisheye global shutter cameras. Different types of hardware acceleration in embedded DSP, VPU or GPU can also be used. Contact us using the form below for more information.

Benchmark results

To demonstrate the performance of the SDK in inside-out tracking, we report the results on the room-scale subset of the TUM Visual-Inertial Dataset

TUM VI Room benchmark results. The error metric is SE3-aligned ATE. SDK version 1.37
Error	1	2	3	4	5	6	Mean
	Sequence
RMSE	1.6 cm	0.9 cm	1.5 cm	1.9 cm	1.9 cm	1.2 cm	1.5 cm
Median	1.4 cm	0.7 cm	1.2 cm	1.7 cm	1.7 cm	1.0 cm	1.3 cm

The benchmark data is used under the Creative Commons 4.0 Attribution License (CC BY 4.0). Please visit the dataset home page or refer to the original publication (Schubert et. al., 2018) for details.