GameWAM ViZDoom
Yunncheng/gamewam-vizdoom on Hugging Face ↗
First-person Doom episodes played by trained agents across four scenarios, with every action and the player state per frame.
- Source
- Hugging Face
- Robot
- Game agent
- Episodes
- 50,000
- Frames
- 79703k at 35 fps
- Tasks
- 4
- Cameras
- 1 first-person
- License
- Apache 2.0
- Export as
- LeRobot · RLDS · MCAP
Samples
What actions do to this data — each one is a run on the same footage.
No sample runs on this dataset yet. Ask for one and we'll run your pick of actions on a few episodes and send you the results.
Request a sample runUse it in an action
ExploreSearch and browse your episodes by what happens in them.DiffusionRe-shoot one capture into new scenes, light and embodiments.LabelDepth, masks, boxes and hands on every frame, automatically.EvaluateTrain and score robot policies on your episodes.
We import this dataset for you: running an action opens a task, and the Action Agent and a reviewer send results back in 12–24 hours.