Sports datasets

A helmet camera on whitewater. Both hands, the paddle and the bow stay in view at close range, and the whole image swings with the wearer's head. Head direction is not gaze: where the eyes pointed needs eye-tracking cameras.

Camera

One worn camera or a camera fixed at the driver’s eye line.

View

Hands, nearby objects and other people from the wearer’s viewpoint.

Useful for

First-person action and hand-object interaction.

Request a sample of egocentric video

A tripod camera sees the whole rider and bike come down rock ledges: body position, bike path and ground in one frame, which a camera on the rider's helmet cannot show. Where the rider looked is not recorded.

Cameras

One or more fixed cameras around a court, pitch, track or training area.

Timing

Timing and calibration are checked against the requested tolerance.

Useful for

Whole-body movement, participant tracking and activity context.

Request a sample of exocentric video

Two global-shutter cameras 109.5 mm apart on a cyclist's helmet, shown after ONME rectified them with the published calibration, on a daytime ride along campus paths. Compare objects at different distances: the railing beside the path shifts more between the two views than the building facade, and that shift is what gives depth. The bicycle and the rider are out of view; the sport shows in the motion.

Cameras

Two synchronised cameras mounted a small distance apart on one wearer.

View

Overlapping first-person images with calibration for stereo geometry.

Useful for

Depth estimation near the wearer and hand-object interaction.

Request a sample of stereo egocentric video

Two wearers exit the same aircraft on the same pass, each with a camera, aligned after capture from a shared exit call. At the door one camera sees the other wearer, at the moment that wearer's own camera shows the door. Wearer B's view is a 360 degree camera reframed to portrait before publication, so it mostly looks back at its wearer. Markings and faces are obscured.

Wearers

Two or more participants record the same session from their own viewpoints.

Timing

Matching timestamps and wearer IDs align every first-person stream.

Useful for

Comparing what participants could see during the same action.

Request a sample of multi-egocentric video

Cameras

A worn camera and external cameras record the same drill or activity.

Timing

The first-person and external streams will share measured synchronisation.

Useful for

Relating a participant’s view to body movement and activity context.

Frames from a helmet camera on a track cycling ride on the left, and on the right the camera path and sparse 3D points ONME reconstructed from them, seen from above and appearing as they are first seen; the clip opens 3 s into the excerpt, with points already drawn. Points are coloured by height above the infield floor, from dark blue at 0 m to orange at 9 m and above, and the roof is hidden. Riders ahead are left out because they move. This is a reconstruction, not a simulation: it has no surfaces and no physics.

Inputs

Recorded camera paths, scene observations and human motion.

Output

Reconstructed geometry or a simulation prepared for an agreed system.

Useful for

Repeatable testing, new viewpoints and synthetic task variation.