Capture catalogue · capture brief
Game Environment & Simulator Data
This brief is under evaluation and not open for commission.
What the brief covers
The brief specifies humanoid agents traversing structured worlds built in Gazebo or MuJoCo, rendered rather than recorded. It is scoped for reinforcement-learning path planning and for validating spatial models against values that are exact by construction. Each frame pairs RGB with a depth matrix, at 60 frames a second.
The specification
| Capture device | Simulator Engine |
|---|---|
| Camera | Synthetic RGB-D @ 60 FPS |
| Spatial tracks | Ground-truth joint angles, SLAM trajectory, and depth matrices |
| Annotation | Tier 1 · Tier 2 |
| Availability | Research phase |
What a delivery carries
Where a recorded brief derives hand pose from video, this one would read the agent's state out of the simulator: the angle of every joint, the trajectory, and the depth itself. The specification adds ground-truth physics states and stereo depth maps to those multi-joint kinematics.
What to know before you commission this
- This brief is synthetic. There is no camera wearer, so the consent chain that covers recorded footage does not apply here.
- It is being evaluated against the standard the rest of the catalogue is held to.
- The fields shown are the proposed delivery shape for this brief.
Sample
Published sample: none