EgoGaming Data › VR gameplay data vs. egocentric video
Comparison
VR Gameplay Data vs. Egocentric Head-Camera Video
Both are first-person data, but they trade off differently. Head-camera video shows the real world, and the wearer’s motion and hand positions usually have to be estimated from it. VR gameplay shows a rendered game, but the head and hand motion is measured by the headset on every frame, together with the player’s inputs. Choose by what your model must learn: real-world appearance, or action and response.
Side by Side
| VR gameplay recordings (EgoGaming) | Egocentric head-camera video | |
|---|---|---|
| Picture | The game’s rendered view, exactly what the player saw | Real-world camera footage |
| Head motion | Measured 6DoF headset pose on every frame | Often absent; some rigs add IMU or SLAM trajectories |
| Hand motion | Measured 6DoF controller poses; hand joints where tracked | Usually estimated from the video afterwards |
| Actions and inputs | Buttons, triggers and sticks where the capture route records them | Implicit in the video |
| Sync | Video, poses and inputs on one capture clock | Depends on the rig |
| Frame rate | The game’s own, about 70–90 fps | Typically 30 or 60 fps |
| Tracking quality | Valid and tracked flags on every reading | Varies; often not reported |
| Bystanders | Other players appear only as avatars in multiplayer games | Real people and places can appear on camera |
| Domain gap | Rendered graphics, game physics | None for real-world tasks |
When VR Gameplay Data Is the Better Fit
- World models of games and interactive 3D worlds, where you need the action behind every frame.
- VR products: motion prediction, bots that move like people, scoring and comfort research.
- Any model where measured, frame-accurate head and hand motion matters more than photorealism.
When Head-Camera Video Is the Better Fit
- Models that must recognize real objects, places and people.
- Robot perception for real environments.
- Tasks that need real physics, contact and forces in the picture.
Many teams use both: real video for appearance, VR data for measured motion and action. If you need human motion for real-world tasks, custom task recording can narrow the gap.
Consent and Identifiability
Both kinds of data can identify people. Head-camera video can show faces and places; VR movement itself can identify a person. EgoGaming players are paid adults who agreed in writing to licensing, each session has a consent record, and the license forbids any attempt to identify them. Read how consent and withdrawal work.
Compare With Your Own Data
The free sample has 100 action clips in RLDS and WebDataset, with the data card.