Per-frame metric scene depth (Depth-Anything-V2 Metric Indoor Large, 20 m ceiling) for the episodes inpainted in VR-VLA/VR-egodex-inpaint-cache-full. Depth is the second input Eq. 4 compositing needs beside the inpainted plate; together they let scripts/egodex/buildtask.py retarget an episode onto all 16 embodiments. Archive names are IDENTICAL to the inpaint repo, one tar per /: Episode numbers match the source dataset's episode%06d.mp4. is part of the key: most task names exist in both train/ and test/ with overlapping episode numbers. Archives are uncompressed tar (PNG payload). train/addremovelidv2.1.tar (308 GB, 1,158 episodes) exceeds the Hub's 50 GB file limit and is stored as…
Organization
VR-VLA
VR-VLA
Models in Library0
Datasets in Library2
Models on Hugging Face7
Followers7
Compare the published evidence
This profile connects the 2 model and dataset records shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Published tasks: Robotics.
Reported licenses: other. Open a record to review its specific license and conditions.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.
Datasets
EgoDex Retarget — 16 robots EgoDex egocentric human demonstrations retargeted onto 16 robot embodiments. For every episode, the human wrist and hand motion is solved into a joint trajectory for each robot, and the robot is rendered and composited into the original scene, over hand-erased frames and per-frame metric depth. Each (robot, task) pair is one LeRobot v3.0 dataset. VR-VLA/VR-egodex-inpaint-cache-full (hand-erased frames), VR-VLA/VR-egodex-depth-cache-full