Upload 82 clip(s) (3.71 GB) — -821U1ZYZOs_245.0_249.0, -821U1ZYZOs_321.4_325.4, -821U1ZYZOs_326.1_330.3, ... (+79 more) [2026-05-22T09:07:28+00:00] (6/82)

Upload 105 clip(s) (4.85 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+102 more) [2026-05-22T07:52:33+00:00] (29/105)

Upload 82 clip(s) (3.72 GB) — -7uLbtVsmQs_874.8_879.1, -7uLbtVsmQs_874.8_879.9, -821U1ZYZOs_245.0_249.0, ... (+79 more) [2026-05-19T10:23:31+00:00] (8/82)

Upload 105 clip(s) (4.88 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+102 more) [2026-05-19T09:13:29+00:00] (30/105)

Upload 105 clip(s) (4.88 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+102 more) [2026-05-19T07:26:37+00:00] (30/105)

Upload 105 clip(s) (4.97 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+102 more) [2026-05-19T03:41:50+00:00] (30/105)

Upload 105 clip(s) (4.97 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+102 more) [2026-05-18T07:24:56+00:00] (30/105)

Upload 105 clip(s) (4.97 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+102 more) [2026-05-17T18:46:58+00:00] (30/105)

Upload 64 clip(s) (3.00 GB) — -7uLbtVsmQs_874.8_879.9, -821U1ZYZOs_245.0_249.0, -821U1ZYZOs_321.4_325.4, ... (+61 more) [2026-05-17T18:02:06+00:00] (7/64)

2026-05-17: scale_sanity Plan A rollout — 33 clips, 61 oid scale overrides (mask×depth physics check, threshold 1.8×) (9/34)

2026-05-15: hand+object Z-axis Gaussian smooth (σ=5 frames, MAD outlier reject) (2/8)

2026-05-15: state machine v2 — STATIC / GRASPED_L/R/BOTH / MOVING + L/R-aware grasp (30/105)

2026-05-15: refresh_hand_scale Procrustes fit on 7 close-up clips (BBQ family + xWBZ) (2/7)

Upload 112 clip(s) (4.99 GB) — --oo8_XIuOM_799.5_809.8, --oo8_XIuOM_900.3_917.4, -0RheyDV3a0_474.8_487.3, ... (+109 more) [2026-05-12T20:49:27+00:00] (30/112)

v1.11: pure canonical_scale (hf_reference simplified). No ICP / Umeyama / sanity refinement. Mesh scale = SAM3D canonical_scale directly; R = canonical_rotation_quat; per-frame translation = obs centroid + Gaussian smooth.

v1.10: restore original 5-stage tracking (estimate_anchor_pose FGR+ICP+Umeyama + track_6dof flow propagation). 40/225 oids get full 6DoF tracking; rest fall back to centroid translation + identity rotation. Scale closer to A100 reference (cereal box 0.219 vs reference 0.263).

v1.9: scale_correction is now ABSOLUTE (PLY units→world m, matches A100 reference convention). Robust min-enclose algorithm: per-axis [2,98]% pre-filter + median over axes of per-axis 90% enclose (depth-fringe-resistant)

v1.8: tracking uses SAM3D canonical_rotation_quat (per-clip initial orientation) + min-scale-95% enclose algorithm; object_pose.bin now reads pose_track_info.T_seq (full 6DoF tracked pose, not stale PCA OBB)

v1.7: SAM3D refill round 2 (226 PLYs, +25 from v1.6) — decoder cleanup frees 1.7GB GPU; native-res bbox crop (SAM3D internally resizes to 518)

Replace samples/ with 21-clip Step 1 test batch (no SAM3D, no 6DoF)

Refresh dataset: 48 trim-reviewed clips, manual SAM3 prompts, full descriptions in scene.json