Sep 21 30 days ago 23:38 Publish Track 2 challenge tiers with support surfaces Replace the previous Track 2 variants with the reviewed challenge release. Publish Tier 1 multiview-caption trajectories and Tier 2 synthetic-noise trajectories, each with one generic monocular exo_camera stream, ground-plane metadata, and one support-surface USDA per episode. Use the reviewed alternate camera for six sequences, remove extra camera streams and source-camera/extrinsic metadata, remove precomputed CARI4D reconstruction results, and omit Tier 2 generation-provenance details. 23:22 Add Track 3 evaluation videos 22:42 Publish reviewed Track 2 challenge tiers (#6 ) zhuxh-nvidia +1: nv-shalinj - Publish reviewed Track 2 challenge tiers (c409ed8ab1d52ee86c279835ba8ea9d54091aa5a)
Co-authored-by: Shalin <nv-shalinj@users.noreply.huggingface.co> 22:22 Update README.md 21:53 Replace Track 1 with 30 selected single-view sequences (#5 ) zhuxh-nvidia +1: dzou-nv - Replace Track 1 with 30 selected single-view sequences (97510bc2c901548eb541f48e183a6e182c56d5cc)
Co-authored-by: Daniel Zou <dzou-nv@users.noreply.huggingface.co> 19:46 Update track_3/v2d-challenge-track3.md 03:14 Organize Track 2 trajectory variants (#4 ) zhuxh-nvidia +1: nv-shalinj - Organize Track 2 trajectory variants (ac080f5fe77f31edb648f641951932cbce24240d)
- Restore Track 2 variant index (a025aff2f0d0cbf156fc07d1bd8b4ade8ab4fffe)
- Align CARI4D trajectories to the first frame (b3d2e7037c417405925ae4b9afac8c87b7ec577b)
- Fix CARI4D object alignment coordinates (2866f79a8d236492e64c95872dfe5c38cca6633f)
- Calibrate synthetic trajectory noise (1e6ab9e12b8842bb9b3d26b1df0c4a44e52c8371)
- Add fitted ground planes to Track 2 (89296f66bd98415f7edb47e96875d2c6ac6063e3)
Co-authored-by: Shalin <nv-shalinj@users.noreply.huggingface.co> Sep 20 23:40 Update README.md Sep 18 22:35 License Track 3 dataset and assets under CC BY-4.0 Sep 17 22:31 Remove confidential Track 3 holdout from public release 21:04 Remove Track 1 ground-truth annotations (#3 ) zhuxh-nvidia +1: dzou-nv - Remove Track 1 ground-truth annotations (290ba28b5ed5cb589a3b600ad54df83105ba3da0)
Co-authored-by: Daniel Zou <dzou-nv@users.noreply.huggingface.co> Sep 10 20:33 updated track3 datacard Sep 2 17:53 Remove Track 3 exocentric videos 07:56 Add track_1: single-view LeRobot v2.1 dataset (67 episodes, 48846 frames, 33 object meshes, SOMA-X + MHR human params) Sep 1 23:21 Move episode 42 to Track 3 public split 23:05 Split Track 3 into public and confidential holdout sets 22:10 dropping episode_000029 21:38 adding URDF 21:25 Updating assets 00:39 Add Track 2 whole-body reconstruction dataset (#2 ) zhuxh-nvidia +1: nv-shalinj - Add Track 2 whole-body reconstruction dataset (0ea7577ec04d29b6bab0f9549d2c51a2bd3fd54d)
- Simplify Track 2 README (723ece5d24e8a679c9085b6c65f4b407bae416db)
- Align Track 2 README with Track 3 style (ffbb62aa4615aefe6623321404a6b11fbb21400b)
Co-authored-by: Shalin <nv-shalinj@users.noreply.huggingface.co> Aug 28 09:58 Update track_3 README 09:55 Add track_3 README 09:51 Add track_3: LeRobot v2.1 dataset (46 episodes, 276 videos, 11 object meshes) (part 2) 09:50 Add track_3: LeRobot v2.1 dataset (46 episodes, 276 videos, 11 object meshes) 09:42 initial commit 
