Initial upload: 164,390 strict-pass clips with retracked tracking + 235B VLM verdicts

README.md

@@ -0,0 +1,128 @@
+---
+language:
+  - en
+license: cc-by-4.0
+task_categories:
+  - robotics
+tags:
+  - pose-estimation
+  - 3d-pose
+  - joint-angles
+  - humanoid
+  - kinetics-700
+  - VLA
+  - filtered
+size_categories:
+  - 100K<n<1M
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: "train-*.parquet"
+---
+
+# movement-strict-164: High-Quality Filtered Pose Dataset
+
+164,390 clips from Kinetics-700 that pass **both** programmatic continuity checks **and** a 235B-parameter Vision-Language Model judge that evaluated the rendered skeleton overlay against the action label. Roughly 60% of clips have been re-tracked through a dedicated multi-frame YOLO + Qwen oracle + sticky IoU tracker pipeline before judgment, replacing the original tracking with a cleaner result.
+
+This is a filtered subset of [`maxsegan/movenet-332`](https://huggingface.co/datasets/maxsegan/movenet-332) intended as the highest-quality tier for training Vision-Language-Action models for humanoid robots. The companion broader release with the same retracking but a more permissive quality bar is at [`maxsegan/movement-287`](https://huggingface.co/datasets/maxsegan/movement-287).
+
+## What "strict" means
+
+Each clip's tracking was re-judged with `Qwen3-VL-235B-A22B-Thinking-FP8` looking at six rendered overlay frames per clip. A clip is **strict** if all of these hold:
+
+- `v2_prog_good` — passes continuity / motion / size checks (no hard cuts, no IoU breaks, no tracking jumps, peak joint speed ≥ 3 m/s, bbox area ≥ 8% of frame, etc.)
+- `v2_vlm_tracking_good` — VLM confirms the same person is cleanly tracked throughout
+- `v2_vlm_motion_matches` — VLM confirms the tracked person is performing the labeled action
+
+This drops slow legitimate motion (texting, archaeological excavation) along with tracking artifacts, in exchange for the cleanest possible velocity profiles and identity continuity. Use the broader `movement-287` set if you want to keep slower actions.
+
+## Re-tracking pipeline
+
+For each clip flagged by either signal in a first pass, the original tracking was replaced via:
+
+1. Multi-frame YOLOv8x candidate pool over 5 sampled frames
+2. `Qwen3-VL` oracle picks the subject on the busiest anchor frame
+3. Sticky IoU tracker (threshold 0.5, size-ratio guard 2.5×, 3-miss tolerance)
+4. Trim to longest valid run containing the anchor; re-detect hard cuts on the trimmed range
+5. Interpolate residual interior NaN bbox gaps
+6. ViTPose-plus-large 2D + MotionAGFormer-B 3D lifting on the new tracking
+
+The retracked clips ship with shorter `num_pose_frames` than the originals when the trimmer cut off bad segments — frame ranges 35–250 are typical.
+
+## Schema
+
+| Column | Type | Description |
+|---|---|---|
+| `clip_id` | string | NPZ filename stem (e.g., `a4UOhPiV4QE_000675_000685`) |
+| `action_class` | string | Kinetics-700 action class name |
+| `youtube_id` | string | YouTube video ID |
+| `time_start`, `time_end` | string | Start/end timestamp from the clip ID |
+| `split` | string | Original `train`/`val` split (preserved from movenet-332) |
+| `instruction` | string | VLM-generated imperative caption |
+| `fps` | float32 | Pose-data FPS (20.0 for retracked clips, original for source=original) |
+| `num_pose_frames` | int32 | Number of pose frames after retrack/trim |
+| `video_width`, `video_height` | int32 | Original video dimensions |
+| `pose3d` | bytes | zlib-compressed float32 `[F, 17, 3]` H36M format |
+| `keypoints2d` | bytes | zlib-compressed float32 `[F, 17, 2]` |
+| `scores2d` | bytes | zlib-compressed float32 `[F, 17]` joint confidences |
+| `bboxes` | bytes | zlib-compressed float32 `[F, 4]` `[x1, y1, x2, y2]` |
+| `joint_angles` | bytes | empty for retracked clips (recompute from `pose3d`); preserved zlib-compressed float32 `[F, 22]` for source=original |
+| `frame_indices` | bytes | zlib-compressed int32 `[F]` mapping pose frame → video frame |
+| `tracking_confidence` | bytes | zlib-compressed float32 `[F]` per-frame mean joint confidence |
+| `has_hard_cuts` | bool | Whether hard cuts were detected in the (possibly retracked) sequence |
+| `quality` | float32 | Original quality score; superseded by `v2_*` columns |
+| `v2_source` | string | `"retrack"` or `"original"` |
+| `v2_prog_good` | bool | Programmatic check verdict |
+| `v2_vlm_tracking_good` | bool | 235B VLM tracking-consistency verdict |
+| `v2_vlm_motion_matches` | bool | 235B VLM motion-matches-label verdict |
+| `v2_vlm_good` | bool | `vlm_tracking_good AND vlm_motion_matches` |
+| `v2_strict` | bool | `prog_good AND vlm_good` — always `True` in this release |
+| `v2_prog_issues` | list[string] | Names of failing programmatic checks (empty in this release) |
+
+Array columns are zlib-compressed raw numpy bytes. Variable-length across clips.
+
+## Loading
+
+```python
+from datasets import load_dataset
+import numpy as np
+import zlib
+
+ds = load_dataset("maxsegan/movement-strict-164", split="train")
+row = ds[0]
+F = row["num_pose_frames"]
+pose3d = np.frombuffer(zlib.decompress(row["pose3d"]),
+                       dtype=np.float32).reshape(F, 17, 3)
+print(row["clip_id"], row["action_class"], row["v2_source"], pose3d.shape)
+```
+
+## Source breakdown (164,390 clips)
+
+| Source | Count | Description |
+|---|---|---|
+| `original` | 107,914 | Phase-1 strict clips that already passed both signals on the original parquet — tracking unchanged from movenet-332 |
+| `retrack` | 56,476 | Phase-2 clips that became strict only after re-tracking |
+
+## Composition stats vs movenet-332
+
+|  | movenet-332 (raw) | movement-strict-164 |
+|---|---|---|
+| Clips | 325,036 | 164,390 |
+| Mean clip duration | ~24 s | ~7 s (after trim-to-valid-run) |
+| Action classes | 700 | 700 |
+| Min clips/class | 12 | varies |
+
+## Skeleton & joint angle conventions
+
+H36M 17-joint skeleton, same as movenet-332. See [`maxsegan/movenet-332`](https://huggingface.co/datasets/maxsegan/movenet-332) for the joint hierarchy and 22-DoF layout.
+
+To recompute joint angles from `pose3d` (the source=retrack rows have empty `joint_angles`), use the helper at `training/kinetics_dataset.py::pose3d_to_joint_angles` in the [GitHub repo](https://github.com/maxsegan/movement).
+
+## Pipeline code
+
+The full re-tracking + judging + filtering pipeline lives at https://github.com/maxsegan/movement (`scripts/retrack_clips.py`, `scripts/scale_judge.py`, `scripts/scale_judge_npz.py`, `scripts/compose_final_dataset.py`, `scripts/package_v2.py`).
+
+## License
+
+Pose annotations and metadata: CC-BY-4.0. Underlying Kinetics-700 videos are subject to their own license; this dataset does not redistribute video content.

train-00000-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b793bca92fe94d4a6d3055cfc9e96be18438f0844772601ecfe236826b1d86a
+size 852837014

train-00001-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbb52e403cb88d85d0ac194b71a9cb13a7c9d9ff55324e7aba78df595d778328
+size 804325822

train-00002-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb248ccd47bdfd3e91b8c4b803d27b18f5a4395e7768d3026d036f769ef5ae82
+size 833410347

train-00003-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc7f269ef0ff578eb20ae3543670c6543e8920203dd4d2c68f2b8aa46eb67f2d
+size 805227445

train-00004-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae94722e4e347120f9c63af955713b3b2a7116761e8de1d0b97cbea2a79e1d57
+size 912616451

train-00005-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d77caf28f72ab024a7b5a6915ed1df0e43ff697c4e9f095cb358c8abf9e5c3
+size 766090137

train-00006-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38b1686743bd3388ce4381cbe75168e2868f3a8f9afc90a181811106f1e7f255
+size 729438434

train-00007-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:472f4c2a97b41ccf9cb71b04bd2f83e3ee52c4c261cd28638bc8434c46902899
+size 810772716

train-00008-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82f859fdda4aae00b348e2c0b38f3121e2805585a8a2438d4603b092be4b5afd
+size 813939793

train-00009-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea199cdf095e49048455fe368665b58f11ef6dc58f7e32a2e16a2f761ea9da25
+size 840430511

train-00010-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35f2edab7a56bdaa698ce3530929c3998ceda81afe9ba37df14e6dc81227b453
+size 794901159

train-00011-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d080dee671037b743cefe25e036cdc2466fb59d6877e43a2b13cf8189225a5b
+size 842032589

train-00012-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:251f4be4bf5c82f2e936f02a6c9a28519408964392af84005195eaa54647b222
+size 893778589

train-00013-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c556c58773db485d6fdf1de44d4e281858369b70094a88a63092278cdeeda542
+size 743533794

train-00014-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:196709ae060e8a9ec335ffa01d6049b99b9a03a886ca309f715a5e368519415b
+size 722578261

train-00015-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed776a9b97905b99d05cb58ff4176c218574824f3e22fa60eb3a81fc8947c0f4
+size 837107952

train-00016-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:827f5bb13af1912ebb3c04792a26b72d9359b6bfd76a1dc59c07b7c90a51dae4
+size 903951849

train-00017-of-00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80225cf674167b1da2ce2618bcda8b682b2993df28ee85f27c23b655fc574731
+size 335996028