IEEE ISMAR 2024
Hand-based motion matching selection for small targets on XR displays.
WASD or the arrow keys.12 targets — museum gallery
Twelve works on a gallery wall, each with its own info marker. The paper's main configuration, put somewhere it would actually be used.
2 closely spaced targets
Two markers on a single artwork, side by side — the hardest case for ray-based pointing, and where motion matching pays off most.
12 targets — plain grid
The original layout with nothing around it: no environment, the original target sizes, and plain hand/joint dropdowns. Use this one to look at timing and correlation on their own.
Every webcam demo takes a ?scene= parameter, or use the Scene dropdown in the controls panel:
museum ·
outdoor ·
plain.
The WebXR versions are still being worked on and may be incomplete or change without notice. The webcam demos above are the stable ones.
12 targets — headset
The plain twelve-target grid driven by the headset's own hand tracking. Enter AR to place it over the room you are actually standing in, or VR for the plain backdrop.
2 targets — headset
The two-target scene driven by the headset's native hand tracking instead of a webcam. Works in AR as well as VR.
12 targets — ray + pinch baseline
The same twelve targets selected the conventional way: point with the headset's hand ray, pinch to confirm. Same scene, same positions, same sizes — only the technique changes, which is what makes it a comparison.
Both headset demos take ?select=ray, so you can run the baseline on
12 targets or
2 targets,
and drop the parameter to switch back to motion matching.
Hand skeleton viewer
Renders every tracked hand joint as a sphere in AR passthrough. Useful for checking what joint data a device actually reports.