Upload README.md with huggingface_hub

README.md

@@ -92,12 +92,13 @@ All repos live under [`MIT-Media-Lab`](https://huggingface.co/MIT-Media-Lab) and
 |---|---:|---:|---|---:|---|
 | **EPIC-KITCHENS-100** | 149,570 | 4,019,534 | episodes + text | 8.70 GB | [`epic30-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/epic30-annotations-v1) |
 | **EgoTouch** | 111,159 | 3,687,389 | episodes + text + **tactile** | 20.12 GB | [`egotouch-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/egotouch-annotations-v1) |
+| **GigaHands** | 70,486 | 2,266,087 | episodes + text | 2.89 GB | [`gigahands-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/gigahands-annotations-v1) |
 | **Ego-Exo4D** | 67,051 | 1,757,474 | text only | 4.09 GB | [`egoexo4d-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/egoexo4d-annotations-v1) |
 | **Something-Something V2** | 52,706 | 1,124,722 | text only | 4.63 GB | [`ssv2-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/ssv2-annotations-v1) |
 | **OakInk2** | 28,264 | 1,371,721 | episodes + text | 1.92 GB | [`oakink2-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/oakink2-annotations-v1) |
 | **TACO** | 23,757 | 736,136 | episodes + text | 1.34 GB | [`taco-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/taco-annotations-v1) |
 | **H2O** | 5,696 | 196,941 | episodes + text | 0.40 GB | [`h2o-annotations-v1`](https://huggingface.co/datasets/MIT-Media-Lab/h2o-annotations-v1) |
-| **total** | **438,203** | **12,893,917** | | **41.2 GB** | |
+| **total** | **508,689** | **15,160,004** | | **44.1 GB** | |
 
 **episodes** = entries in `episode_frame_index.npz`, i.e. what a training run actually sees.
 **training samples** = rows of `index_frame_pair`; the row number *is* the sample id.
@@ -114,6 +115,7 @@ instruction. That is why the published counts are below the totals we cut:
 |---|---:|---:|---:|
 | EPIC-KITCHENS-100 | 151,502 | 149,570 | 1,932 (1.3%) |
 | EgoTouch | 147,386 | 111,159 | 36,227 (24.6%) |
+| GigaHands | 92,365 | 70,486 | 21,879 (23.7%) |
 | Ego-Exo4D | 67,051 | 67,051 | 0 |
 | Something-Something V2 | 52,706 | 52,706 | 0 |
 | OakInk2 | 37,692 | 28,264 | 9,427 (25.0%) |
@@ -126,11 +128,10 @@ slightly below VITRA-1M's published index (52,718 and 67,053) because round 2 ma
 sentences `unusable` and we deleted those episodes: **12** from `ssv2`, **2** from `egoexo4d`.
 
 
-### Every camera here is head-mounted
+### A note on viewpoint
 
-Every dataset above is filmed from the actor's own head or chest. We checked this rather than
-trusting the labels — for each domain we reconstructed the camera centre from the per-frame
-extrinsics and measured how far it travels within a source video:
+For each domain we reconstructed the camera centre from the per-frame extrinsics and measured
+how far it travels within a source video, rather than trusting the labels:
 
 | | camera displacement within a video (median / max, m) |
 |---|---|
@@ -139,13 +140,17 @@ extrinsics and measured how far it travels within a source video:
 | ARCTIC | 0.044 / 0.212 |
 | H2O | 0.022 / 0.149 |
 | TACO | 0.017 / 0.041 |
+| GigaHands | **0.000000 / 0.000000** — static rig |
 
-**DexYCB and GigaHands were removed from this collection on 2026-08-30** for failing that test:
-both are **static third-person rigs**, and their camera displacement is exactly `0.000000` m —
-not "small", but bit-for-bit zero, on every episode. DexYCB is 8 fixed RealSense cameras around
-a table (its own `camera.role` field says `allocentric`), and GigaHands is the BRICS studio ring.
-Neither ships a head-mounted view, so there was nothing to filter down to. Their hand poses and
-instructions are fine; the viewpoint distribution simply is not what these episodes are for.
+Those five are head- or chest-mounted. **GigaHands is not**: it is the BRICS studio ring, a fixed
+third-person camera array, and its reconstructed camera centre is bit-for-bit `0.000000` m on
+every episode. Its hand poses and instructions are fine — only the viewpoint differs — so it is
+kept here and labelled rather than dropped.
+
+**DexYCB was removed from this collection on 2026-08-30.** It is 8 fixed RealSense cameras around
+a table; its own `camera.role` field reads `allocentric` on all 15,878 episodes and the camera
+displacement is exactly zero. Unlike OakInk2, which ships an egocentric view alongside three
+allocentric ones, DexYCB has no head-mounted camera at all, so there was nothing to filter down to.
 
 ### Not published yet