Datasets:

chen-si-02
/

AnyHand-Dataset

@@ -1,271 +1,286 @@
----
-pretty_name: "AnyHand: A Large-Scale Synthetic Dataset for RGB(-D) Hand Pose Estimation"
-license: cc-by-nc-nd-4.0
-size_categories:
-  - 1M<n<10M
-task_categories:
-  - keypoint-detection
-  - depth-estimation
-  - other
-tags:
-  - image
-  - 3d
-  - webdataset
-  - computer-vision
-  - hand-pose-estimation
-  - 3d-hand-reconstruction
-  - hand-object-interaction
-  - rgb-d
-  - synthetic-data
-  - mano
-  - neurips-2026
-  - arxiv:2603.25726
-viewer: false
----
-<div align="center">
-# 🤲 AnyHand
-### A Large-Scale Synthetic Dataset for RGB(-D) Hand Pose Estimation
-**NeurIPS 2026 · Evaluations and Datasets Track**
-[![Project Page](https://img.shields.io/badge/Project-Website-2563eb?style=flat-square)](https://chen-si-cs.github.io/projects/AnyHand/)
-[![Paper](https://img.shields.io/badge/Paper-arXiv%3A2603.25726-b31b1b?style=flat-square)](https://arxiv.org/abs/2603.25726)
-[![Code](https://img.shields.io/badge/Code-GitHub-181717?style=flat-square&logo=github)](https://github.com/chen-si-cs/AnyHand)
-[![Checkpoints](https://img.shields.io/badge/Models-Hugging%20Face-ffcc4d?style=flat-square)](https://huggingface.co/chen-si-02/AnyHand-Models)
-[![Live Demo](https://img.shields.io/badge/Demo-Hugging%20Face%20Space-f97316?style=flat-square)](https://chen-si-02-anyhand-demo.hf.space/)
-**6.6M RGB-D images · 47,438 hand shapes · 10,240 hand textures · 500K+ objects**
-[**Explore the project**](https://chen-si-cs.github.io/projects/AnyHand/) · [**Read the paper**](https://arxiv.org/pdf/2603.25726) · [**Run the Colab demo**](https://colab.research.google.com/drive/1O71DiPWzNlIflCVTViJeXp2VnfffHnTt)
-</div>
-[![AnyHand teaser: diverse poses and qualitative hand-reconstruction results](https://chen-si-cs.github.io/projects/AnyHand/figs/teaser.svg)](https://chen-si-cs.github.io/projects/AnyHand/)
-> **Official dataset release.** This Hugging Face repository is the **canonical distribution location for AnyHand data**. The figures and statistics below describe the **complete 6.6-million-image dataset** reported in the paper; the set of shards currently available in the repository may be smaller while the upload is in progress. Refer to [`manifests/shards.jsonl`](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/blob/main/manifests/shards.jsonl) and [`transfer_summary.json`](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/blob/main/transfer_summary.json) for the published inventory and transfer status. Please do not assume that all 6.6M images are already online.
-## Overview
-**AnyHand** is a large-scale synthetic dataset for **3D hand pose and mesh reconstruction**, supporting both **RGB-only** and **RGB-D** learning. It brings together diverse hand articulations, realistic skin and forearm appearance, varied cameras and illumination, **hand-object occlusion**, and paired geometric supervision at a scale difficult to obtain from real capture.
-The dataset has two complementary branches:
-| Branch | Description | Images | Rendered scenes |
-|:--|:--|--:|--:|
-| **AnyHand-Single** | Isolated hands with attached forearms, broad pose/shape/appearance and viewpoint diversity | **2.5M** | 1.25M |
-| **AnyHand-Interact** | Hands manipulating objects, including natural grasps and object-induced occlusion | **4.1M** | 2.05M |
-| **Total** | Aligned RGB-D and rich geometric annotations | **6.6M** | **3.30M** |
-Each scene is rendered from **two camera viewpoints**. The dataset is designed to **complement real-world training datasets**, not necessarily to replace them: the paper shows that combining synthetic and real data improves hand reconstruction and generalization.
-### At a glance
-| Diversity / supervision | Scale or detail |
-|:--|:--|
-| Hand shapes | **47,438** MANO shape configurations |
-| Hand textures | **10,240** appearances generated using Handy |
-| Forearm textures | **254** textures from SMPLitex |
-| Background sources | **1,270** indoor images and HDR environment maps |
-| Hand-object interactions | **500K+** objects represented in GraspXL / Objaverse |
-| Ground truth and metadata | RGB, depth, foreground masks, bounding boxes, camera parameters, and 3D pose/shape annotations |
-### Visual examples
-**AnyHand-Single — diverse hands, textures, scenes, and viewpoints**
-![Qualitative AnyHand-Single samples](https://chen-si-cs.github.io/projects/AnyHand/figs/single_hand_demo.svg)
-**AnyHand-Interact — varied objects, grasps, and occlusions**
-![Qualitative AnyHand-Interact samples](https://chen-si-cs.github.io/projects/AnyHand/figs/interact_demo.svg)
-More examples, the generation pipeline, and reconstruction comparisons are available on the [project website](https://chen-si-cs.github.io/projects/AnyHand/).
-## Get the data
-The dataset is distributed **directly from this Hugging Face dataset repository** in its original TAR-shard format. You do not need to install a hand reconstruction network or obtain MANO assets just to download or inspect the RGB/depth files.
-### 1. Download a manifest and an example shard
-Install the Hugging Face command-line client:
-```bash
-python -m pip install -U huggingface_hub
-```
-First, get the published shard inventory:
-```bash
-hf download chen-si-02/AnyHand-Dataset \
-  manifests/shards.jsonl \
-  --repo-type dataset \
-  --local-dir ./AnyHand
-```
-Then download **one** shard, replacing the example path with an entry from the manifest if necessary:
-```bash
-hf download chen-si-02/AnyHand-Dataset \
-  data/train/single/000000.tar \
-  --repo-type dataset \
-  --local-dir ./AnyHand
-```
-For a subset, list the precise TAR paths you want. Avoid downloading the full collection unless your storage and network capacity are sufficient.
-### 2. Programmatic access
-Download a selected TAR via `huggingface_hub` and inspect its RGB images with standard Python libraries:
-```python
-from io import BytesIO
-import tarfile
-from huggingface_hub import hf_hub_download
-from PIL import Image
-shard_path = hf_hub_download(
-    repo_id="chen-si-02/AnyHand-Dataset",
-    repo_type="dataset",
-    filename="data/train/single/000000.tar",  # replace using the manifest
-)
-with tarfile.open(shard_path, "r") as archive:
-    rgb_members = [
-        m for m in archive.getmembers()
-        if m.isfile()
-        and m.name.lower().endswith((".jpg", ".jpeg", ".png"))
-        and not m.name.endswith((".depth.png", ".mask.png"))
-    ]
-    member = rgb_members[0]
-    with archive.extractfile(member) as fp:
-        rgb = Image.open(BytesIO(fp.read())).convert("RGB")
-    print(member.name, rgb.size)
-    rgb.show()
-```
-Install Pillow for the example with `python -m pip install pillow`. **Do not unpickle the raw `.data.pyd` labels using an unrestricted loader on untrusted data**; see the annotation notes below.
-<!-- Once the independent toolkit repository is publicly accessible, add its installation
-     and safe-label-reading example here:
-     https://github.com/chen-si-cs/anyhand-toolkit
-     from anyhand_toolkit import Shard
--->
-## Data organization and annotation format
-### Repository layout
-Published files retain their original train/test split grouping **where explicitly present**. Example paths are shown below; refer to the manifest for the actual uploaded files and available splits.
-```text
-AnyHand-Dataset/
-├── README.md
-├── manifests/
-│   └── shards.jsonl
-├── transfer_summary.json
-└── data/
-    └── train/                    # illustrative split
-        ├── single/
-        │   └── 000000.tar
-        ├── interact-indoor/
-        └── interact-outdoor/
-```
-The public-facing category names map to source labels as follows: `dpx` → `single`, `gxl-indoor` → `interact-indoor`, and `gxl-ourdoor` / `gxl-outdoor` → `interact-outdoor`. These are **storage categories**; the scientific branches remain **AnyHand-Single** and **AnyHand-Interact**.
-Each published standard TAR shard currently contains **1,024 paired samples**. Files belonging to one sample share a common basename:
-```text
-00000000.jpg           # RGB image; some shards use PNG/JPEG
-00000000.depth.png     # uint16 depth image
-00000000.mask.png      # optional grayscale foreground mask
-00000000.data.pyd      # legacy NumPy pickle containing [annotation_dict]
-```
-The dataset contains geometric annotations such as 2D/3D hand keypoints, camera information, and hand pose/shape parameters. For the fields explicitly documented in the released TAR reader format:
-| Item | Convention |
-|:--|:--|
-| `keypoints_2d` | **Full-image pixel** coordinates with confidence values |
-| `keypoints_3d` | **Camera-space meters** with confidence values |
-| Camera axes | **+X right**, **+Y down**, **+Z forward** |
-| `cam_t` | Translation is **already reflected in stored 3D joints**; do **not** add it again |
-| `depth_scale_to_m` | Multiply raw `uint16` depth values by this per-sample factor to obtain **meters** |
-| `depth_invalid_code` | Identifies depth pixels to treat as invalid |
-| Confidence | Not a guaranteed or independently verified visibility/occlusion mask |
-**Depth caveat.** Foreground hand/arm depth is produced by rendering. Background depth is estimated using **MoGe-2** and composited in camera space, so the dense depth image is **not uniformly exact ground truth**, particularly near foreground/background boundaries. Foreground masks can be used to focus losses on relevant pixels when available.
-**Security note.** The `.data.pyd` files use a legacy NumPy-pickle representation rather than JSON. Only load such annotations from a trusted source with an appropriately restricted/validated loader; arbitrary pickle loading can execute code. The ordinary image/depth files are accessible without deserializing the labels.
-### Integrity and release tracking
-The publication pipeline validates the RGB, depth, and annotation payloads, checks image readability, checks TAR termination and optional masks where present, and records file metadata. The [`manifests/shards.jsonl`](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/blob/main/manifests/shards.jsonl) inventory contains each published TAR's path, byte size, sample count, SHA-256 digest, and source identifiers. `transfer_summary.json` reports the latest upload/audit status; files that fail validation are excluded rather than published as partially valid shards.
-For large experiments, **pin a repository revision** and retain a copy of the manifest so that experiments use a reproducible set of shards, particularly while uploads are ongoing.
-## Dataset creation
-Our generation pipeline combines the following components:
-1. **Pose and shape sampling.** MANO shapes are sampled from empirical distributions of real hand datasets; articulated poses come from the **DPoser-Hand** diffusion prior.
-2. **Appearance.** **Handy** supplies rich hand textures, with additional appearance variation; **SMPL-H** forearms with **SMPLitex** textures provide realistic wrist/arm continuity.
-3. **Hand-object scenes.** **GraspXL** simulation provides contact-consistent hand-object motions, paired with textured assets from **Objaverse**.
-4. **Rendering and cameras.** **SAPIEN** renders hands and scenes with randomized viewpoints, intrinsics, backgrounds, and lighting; each scene yields two views.
-5. **RGB-D supervision.** Rendered foreground depth and **MoGe-2** background estimates are composed to produce aligned RGB and dense depth, along with geometric labels.
-![AnyHand generation pipeline](https://chen-si-cs.github.io/projects/AnyHand/figs/anyhand_pipeline.svg)
-For the implementation and detailed distributions, please see [Section 3 and Appendix C of the paper](https://arxiv.org/pdf/2603.25726) and the [project page](https://chen-si-cs.github.io/projects/AnyHand/).
-## Results reported in the paper
-In controlled experiments, AnyHand is added to the training data of existing methods while preserving the baseline architectures and training protocols. Representative **PA-MPJPE** results (millimeters; **lower is better**) are:
-| Benchmark | Model | Original training | Training with AnyHand |
-|:--|:--|--:|--:|
-| FreiHAND | HaMeR | 6.000 | **5.545** |
-| HO-3D v2 | WiLoR | 7.500 | **7.355** |
-| HO-Cap *(zero-shot)* | HaMeR | 4.940 | **4.660** |
-The paper also studies RGB-D depth fusion, synthetic-data scaling, and real/synthetic mixing. These numbers are **research results for the configurations reported in the paper**, not precomputed metrics attached to every published TAR shard. See the [paper](https://arxiv.org/abs/2603.25726) for full protocols and additional benchmarks.
-![In-the-wild reconstruction comparisons](https://chen-si-cs.github.io/projects/AnyHand/figs/in-the-wild-vis.svg)
-## Intended use and limitations
-**Intended uses:** research into 3D hand pose/mesh reconstruction, hand-object interactions, RGB-D supervision, synthetic-to-real generalization, and dataset-scaling or data-mixture strategies. The dataset also supports controlled analysis of appearance, pose, and viewpoint diversity.
-**Limitations and considerations:**
-- **Synthetic-to-real gap:** the coverage of rendered appearances, hands, objects, lighting, and behaviors does not fully represent real-world imagery.
-- **Pose realism:** the DPoser-Hand prior and biomechanical filtering improve plausibility but may still admit anatomically unlikely configurations.
-- **Depth quality:** the background portion of a dense depth map is estimated and composited, not measured by a real depth sensor; boundary artifacts are possible.
-- **Training recipe matters:** the paper finds that appropriately **mixing AnyHand with real data** works better than using synthetic samples alone in the studied setup.
-- **Asset and model conditions:** constituent resources (including MANO-related assets, textures, and Objaverse objects) may carry their own terms. Users are responsible for complying with applicable third-party rights and restrictions.
-See [Appendix F](https://arxiv.org/pdf/2603.25726) for the limitations described in the publication.
-## License and citation
-**Dataset license:** [Creative Commons Attribution–NonCommercial–NoDerivatives 4.0 International (CC BY-NC-ND 4.0)](https://creativecommons.org/licenses/by-nc-nd/4.0/). This license applies to the dataset release, and should not be confused with the licenses of code, pretrained models, or third-party materials.
-If you use **AnyHand** in your work, please cite the **NeurIPS 2026 paper**:
-```bibtex
-@inproceedings{si2026anyhand,
-  title     = {{AnyHand}: A Large-Scale Synthetic Dataset for {RGB(-D)} Hand Pose Estimation},
-  author    = {Si, Chen and Liu, Yulin and Ai, Bo and Xie, Jianwen and
-               Potamias, Rolandos Alexandros and Zheng, Chuanxia and Su, Hao},
-  booktitle = {Advances in Neural Information Processing Systems},
-  year      = {2026}
-}
-```
-**Links:** [Project website](https://chen-si-cs.github.io/projects/AnyHand/) · [Paper](https://arxiv.org/abs/2603.25726) · [Code](https://github.com/chen-si-cs/AnyHand) · [Checkpoints](https://huggingface.co/chen-si-02/AnyHand-Models) · [Interactive demo](https://chen-si-02-anyhand-demo.hf.space/) · [Dataset discussions](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/discussions)
-For questions or release issues, please use the [GitHub issue tracker](https://github.com/chen-si-cs/AnyHand/issues) or the [Hugging Face dataset discussions](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/discussions).

+---
+pretty_name: "AnyHand: A Large-Scale Synthetic Dataset for RGB(-D) Hand Pose Estimation"
+license: cc-by-nc-nd-4.0
+size_categories:
+  - 1M<n<10M
+task_categories:
+  - keypoint-detection
+  - depth-estimation
+  - other
+tags:
+  - image
+  - 3d
+  - webdataset
+  - computer-vision
+  - hand-pose-estimation
+  - 3d-hand-reconstruction
+  - hand-object-interaction
+  - rgb-d
+  - synthetic-data
+  - mano
+  - neurips-2026
+  - arxiv:2603.25726
+viewer: false
+---
+<div align="center">
+# 🤲 AnyHand
+### A Large-Scale Synthetic Dataset for RGB(-D) Hand Pose Estimation
+**NeurIPS 2026 · Evaluations and Datasets Track**
+[![Project Page](https://img.shields.io/badge/Project-Website-2563eb?style=flat-square)](https://chen-si-cs.github.io/projects/AnyHand/)
+[![Paper](https://img.shields.io/badge/Paper-arXiv%3A2603.25726-b31b1b?style=flat-square)](https://arxiv.org/abs/2603.25726)
+[![Code](https://img.shields.io/badge/Code-GitHub-181717?style=flat-square&logo=github)](https://github.com/chen-si-cs/AnyHand)
+[![Checkpoints](https://img.shields.io/badge/Models-Hugging%20Face-ffcc4d?style=flat-square)](https://huggingface.co/chen-si-02/AnyHand-Models)
+[![Live Demo](https://img.shields.io/badge/Demo-Hugging%20Face%20Space-f97316?style=flat-square)](https://chen-si-02-anyhand-demo.hf.space/)
+**6,304,768 released RGB-D samples · 6,157 TAR shards · 1,024 samples per shard**
+[**Explore the project**](https://chen-si-cs.github.io/projects/AnyHand/) · [**Read the paper**](https://arxiv.org/pdf/2603.25726) · [**Run the Colab demo**](https://colab.research.google.com/drive/1O71DiPWzNlIflCVTViJeXp2VnfffHnTt)
+</div>
+[![AnyHand teaser: diverse poses and qualitative hand-reconstruction results](https://chen-si-cs.github.io/projects/AnyHand/figs/teaser.svg)](https://chen-si-cs.github.io/projects/AnyHand/)
+> **Official dataset release.** The completed transfer contains **6,157 standard TARs / 6,304,768 samples**, covering all **997 bundles in the pinned Harvard Dataverse source catalogue**. HF file sizes and SHA-256 metadata were verified against the manifest. The catalogue was captured from `doi:10.7910/DVN/VBYEK3`, version `:draft`; source identifiers remain in the manifest. The current live Dataverse inventory was not rechecked in this audit.
+## Published inventory
+| Folder | Standard TARs | Samples | Numeric IDs |
+|:--|--:|--:|:--|
+| `data/train/single` | 2,440 | 2,498,560 | `000000–002439`, consecutive |
+| `data/train/interact-env` | 1,630 | 1,669,120 | `000000–001919`, with 290 absent IDs |
+| `data/train/interact-indoor` | 2,087 | 2,137,088 | `000000–002086`, consecutive numeric suffixes |
+| **Total** | **6,157** | **6,304,768** | Training shards |
+**Numbering gaps:** the environment-interaction series lacks five IDs every 25: `000470–000474`, `000495–000499`, ..., `001895–001899` (58 ranges / 290 TAR IDs). These IDs are also absent from the pinned source catalogue; every catalogued bundle's numeric member range is represented on HF. They are not missing transfers relative to that catalogue. If each missing ID denotes a 1,024-sample shard, those IDs would account for 296,960 additional samples, bringing the numbered inventory to 6,601,728. Their contents and availability have not been verified, so they are not counted as released data.
+The indoor series uses `000000.tar–001089.tar`, followed by `indoor_001090.tar–indoor_002086.tar`. The source filenames and TAR contents are preserved. Use manifest paths rather than constructing filenames by assuming a constant prefix or a gap-free sequence.
+See [`manifests/shards.jsonl`](manifests/shards.jsonl), [`transfer_summary.json`](transfer_summary.json), [`release_audit.json`](release_audit.json), and the [complete absent-ID list](manifests/missing_tar_ids.txt). The audit checks remote size/SHA-256 metadata at its recorded immutable revision; it does not re-download and decode every TAR. Sample counts come from the transfer's previously validated manifest.
+## Overview
+**AnyHand** is a large-scale synthetic dataset for **3D hand pose and mesh reconstruction**, supporting both **RGB-only** and **RGB-D** learning. It brings together diverse hand articulations, realistic skin and forearm appearance, varied cameras and illumination, **hand-object occlusion**, and paired geometric supervision at a scale difficult to obtain from real capture.
+The paper describes two complementary branches. **The following figures are the scale reported in the paper, not the exact current HF inventory; use the published-inventory table above for available samples.**
+| Branch | Description | Images | Rendered scenes |
+|:--|:--|--:|--:|
+| **AnyHand-Single** | Isolated hands with attached forearms, broad pose/shape/appearance and viewpoint diversity | **2.5M** | 1.25M |
+| **AnyHand-Interact** | Hands manipulating objects, including natural grasps and object-induced occlusion | **4.1M** | 2.05M |
+| **Total** | Aligned RGB-D and rich geometric annotations | **6.6M** | **3.30M** |
+Each scene is rendered from **two camera viewpoints**. The dataset is designed to **complement real-world training datasets**, not necessarily to replace them: the paper shows that combining synthetic and real data improves hand reconstruction and generalization.
+### At a glance
+| Diversity / supervision | Scale or detail |
+|:--|:--|
+| Hand shapes | **47,438** MANO shape configurations |
+| Hand textures | **10,240** appearances generated using Handy |
+| Forearm textures | **254** textures from SMPLitex |
+| Background sources | **1,270** indoor images and HDR environment maps |
+| Hand-object interactions | **500K+** objects represented in GraspXL / Objaverse |
+| Ground truth and metadata | RGB, depth, foreground masks, bounding boxes, camera parameters, and 3D pose/shape annotations |
+### Visual examples
+**AnyHand-Single — diverse hands, textures, scenes, and viewpoints**
+![Qualitative AnyHand-Single samples](https://chen-si-cs.github.io/projects/AnyHand/figs/single_hand_demo.svg)
+**AnyHand-Interact — varied objects, grasps, and occlusions**
+![Qualitative AnyHand-Interact samples](https://chen-si-cs.github.io/projects/AnyHand/figs/interact_demo.svg)
+More examples, the generation pipeline, and reconstruction comparisons are available on the [project website](https://chen-si-cs.github.io/projects/AnyHand/).
+## Get the data
+The dataset is distributed **directly from this Hugging Face dataset repository** in its original TAR-shard format. You do not need to install a hand reconstruction network or obtain MANO assets just to download or inspect the RGB/depth files.
+### 1. Download a manifest and an example shard
+This public repository has automatic access gating. Request access on the dataset page, accept the release conditions, and authenticate with an account that has access:
+```bash
+python -m pip install -U huggingface_hub
+hf auth login
+```
+For Python downloads, the HF client uses your locally saved token; do not put tokens in source code.
+First, get the published shard inventory:
+```bash
+hf download chen-si-02/AnyHand-Dataset \
+  manifests/shards.jsonl \
+  --repo-type dataset \
+  --local-dir ./AnyHand
+```
+Then download **one** shard, replacing the example path with an entry from the manifest if necessary:
+```bash
+hf download chen-si-02/AnyHand-Dataset \
+  data/train/single/000000.tar \
+  --repo-type dataset \
+  --local-dir ./AnyHand
+```
+For a subset, list the precise TAR paths you want. Avoid downloading the full collection unless your storage and network capacity are sufficient.
+### 2. Programmatic access
+Download a selected TAR via `huggingface_hub` and inspect its RGB images with standard Python libraries:
+```python
+from io import BytesIO
+import tarfile
+from huggingface_hub import hf_hub_download
+from PIL import Image
+shard_path = hf_hub_download(
+    repo_id="chen-si-02/AnyHand-Dataset",
+    repo_type="dataset",
+    filename="data/train/single/000000.tar",  # replace using the manifest
+)
+with tarfile.open(shard_path, "r") as archive:
+    rgb_members = [
+        m for m in archive.getmembers()
+        if m.isfile()
+        and m.name.lower().endswith((".jpg", ".jpeg", ".png"))
+        and not m.name.endswith((".depth.png", ".mask.png"))
+    ]
+    member = rgb_members[0]
+    with archive.extractfile(member) as fp:
+        rgb = Image.open(BytesIO(fp.read())).convert("RGB")
+    print(member.name, rgb.size)
+    rgb.show()
+```
+Install Pillow for the example with `python -m pip install pillow`. **Do not unpickle the raw `.data.pyd` labels using an unrestricted loader on untrusted data**; see the annotation notes below.
+## Data organization and annotation format
+### Repository layout
+Published files retain their original train/test split grouping **where explicitly present**. This release contains the training folders below; the manifest gives the exact available paths.
+```text
+AnyHand-Dataset/
+├── README.md
+├── manifests/
+│   └── shards.jsonl
+├── transfer_summary.json
+├── release_audit.json
+└── data/
+    └── train/
+        ├── single/
+        │   └── 000000.tar
+        ├── interact-env/
+        └── interact-indoor/
+```
+The published category names map to source labels as follows: `dpx` → `single`, `graspxl_env` → `interact-env`, and `graspxl_indoor` → `interact-indoor`. The toolkit also recognizes the earlier `gxl-indoor` → `interact-indoor` and `gxl-ourdoor` / `gxl-outdoor` → `interact-outdoor` aliases; this inventory has no `interact-outdoor` folder. These are **storage categories**; the scientific branches remain **AnyHand-Single** and **AnyHand-Interact**.
+Each published standard TAR shard currently contains **1,024 paired samples**. Files belonging to one sample share a common basename:
+```text
+00000000.jpg           # RGB image; some shards use PNG/JPEG
+00000000.depth.png     # uint16 depth image
+00000000.mask.png      # optional grayscale foreground mask
+00000000.data.pyd      # legacy NumPy pickle containing [annotation_dict]
+```
+The dataset contains geometric annotations such as 2D/3D hand keypoints, camera information, and hand pose/shape parameters. For the fields explicitly documented in the released TAR reader format:
+| Item | Convention |
+|:--|:--|
+| `keypoints_2d` | **Full-image pixel** coordinates with confidence values |
+| `keypoints_3d` | **Camera-space meters** with confidence values |
+| Camera axes | **+X right**, **+Y down**, **+Z forward** |
+| `cam_t` | Translation is **already reflected in stored 3D joints**; do **not** add it again |
+| `depth_scale_to_m` | Multiply raw `uint16` depth values by this per-sample factor to obtain **meters** |
+| `depth_invalid_code` | Identifies depth pixels to treat as invalid |
+| Confidence | Not a guaranteed or independently verified visibility/occlusion mask |
+**Depth caveat.** Foreground hand/arm depth is produced by rendering. Background depth is estimated using **MoGe-2** and composited in camera space, so the dense depth image is **not uniformly exact ground truth**, particularly near foreground/background boundaries. Foreground masks can be used to focus losses on relevant pixels when available.
+**Security note.** The `.data.pyd` files use a legacy NumPy-pickle representation rather than JSON. Only load such annotations from a trusted source with an appropriately restricted/validated loader; arbitrary pickle loading can execute code. The ordinary image/depth files are accessible without deserializing the labels.
+### Integrity and release tracking
+The publication pipeline validates the RGB, depth, and annotation payloads, checks image readability, checks TAR termination and optional masks where present, and records file metadata. The [`manifests/shards.jsonl`](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/blob/main/manifests/shards.jsonl) inventory contains each published TAR's path, byte size, sample count, SHA-256 digest, and source identifiers. `transfer_summary.json` reports the latest upload/audit status; files that fail validation are excluded rather than published as partially valid shards.
+For large experiments, **pin a repository revision** and retain a copy of the manifest so that experiments use a reproducible set of shards, when the repository is updated.
+## Dataset creation
+Our generation pipeline combines the following components:
+1. **Pose and shape sampling.** MANO shapes are sampled from empirical distributions of real hand datasets; articulated poses come from the **DPoser-Hand** diffusion prior.
+2. **Appearance.** **Handy** supplies rich hand textures, with additional appearance variation; **SMPL-H** forearms with **SMPLitex** textures provide realistic wrist/arm continuity.
+3. **Hand-object scenes.** **GraspXL** simulation provides contact-consistent hand-object motions, paired with textured assets from **Objaverse**.
+4. **Rendering and cameras.** **SAPIEN** renders hands and scenes with randomized viewpoints, intrinsics, backgrounds, and lighting; each scene yields two views.
+5. **RGB-D supervision.** Rendered foreground depth and **MoGe-2** background estimates are composed to produce aligned RGB and dense depth, along with geometric labels.
+![AnyHand generation pipeline](https://chen-si-cs.github.io/projects/AnyHand/figs/anyhand_pipeline.svg)
+For the implementation and detailed distributions, please see [Section 3 and Appendix C of the paper](https://arxiv.org/pdf/2603.25726) and the [project page](https://chen-si-cs.github.io/projects/AnyHand/).
+## Results reported in the paper
+In controlled experiments, AnyHand is added to the training data of existing methods while preserving the baseline architectures and training protocols. Representative **PA-MPJPE** results (millimeters; **lower is better**) are:
+| Benchmark | Model | Original training | Training with AnyHand |
+|:--|:--|--:|--:|
+| FreiHAND | HaMeR | 6.000 | **5.545** |
+| HO-3D v2 | WiLoR | 7.500 | **7.355** |
+| HO-Cap *(zero-shot)* | HaMeR | 4.940 | **4.660** |
+The paper also studies RGB-D depth fusion, synthetic-data scaling, and real/synthetic mixing. These numbers are **research results for the configurations reported in the paper**, not precomputed metrics attached to every published TAR shard. See the [paper](https://arxiv.org/abs/2603.25726) for full protocols and additional benchmarks.
+![In-the-wild reconstruction comparisons](https://chen-si-cs.github.io/projects/AnyHand/figs/in-the-wild-vis.svg)
+## Intended use and limitations
+**Intended uses:** research into 3D hand pose/mesh reconstruction, hand-object interactions, RGB-D supervision, synthetic-to-real generalization, and dataset-scaling or data-mixture strategies. The dataset also supports controlled analysis of appearance, pose, and viewpoint diversity.
+**Limitations and considerations:**
+- **Synthetic-to-real gap:** the coverage of rendered appearances, hands, objects, lighting, and behaviors does not fully represent real-world imagery.
+- **Pose realism:** the DPoser-Hand prior and biomechanical filtering improve plausibility but may still admit anatomically unlikely configurations.
+- **Depth quality:** the background portion of a dense depth map is estimated and composited, not measured by a real depth sensor; boundary artifacts are possible.
+- **Training recipe matters:** the paper finds that appropriately **mixing AnyHand with real data** works better than using synthetic samples alone in the studied setup.
+- **Asset and model conditions:** constituent resources (including MANO-related assets, textures, and Objaverse objects) may carry their own terms. Users are responsible for complying with applicable third-party rights and restrictions.
+See [Appendix F](https://arxiv.org/pdf/2603.25726) for the limitations described in the publication.
+## License and citation
+**Dataset license:** [Creative Commons Attribution–NonCommercial–NoDerivatives 4.0 International (CC BY-NC-ND 4.0)](https://creativecommons.org/licenses/by-nc-nd/4.0/). This license applies to the dataset release, and should not be confused with the licenses of code, pretrained models, or third-party materials.
+If you use **AnyHand** in your work, please cite the **NeurIPS 2026 paper**:
+```bibtex
+@inproceedings{si2026anyhand,
+  title     = {{AnyHand}: A Large-Scale Synthetic Dataset for {RGB(-D)} Hand Pose Estimation},
+  author    = {Si, Chen and Liu, Yulin and Ai, Bo and Xie, Jianwen and
+               Potamias, Rolandos Alexandros and Zheng, Chuanxia and Su, Hao},
+  booktitle = {Advances in Neural Information Processing Systems},
+  year      = {2026}
+}
+```
+**Links:** [Project website](https://chen-si-cs.github.io/projects/AnyHand/) · [Paper](https://arxiv.org/abs/2603.25726) · [Code](https://github.com/chen-si-cs/AnyHand) · [Checkpoints](https://huggingface.co/chen-si-02/AnyHand-Models) · [Interactive demo](https://chen-si-02-anyhand-demo.hf.space/) · [Dataset discussions](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/discussions)
+For questions or release issues, please use the [GitHub issue tracker](https://github.com/chen-si-cs/AnyHand/issues) or the [Hugging Face dataset discussions](https://huggingface.co/datasets/chen-si-02/AnyHand-Dataset/discussions).

manifests/missing_tar_ids.txt ADDED Viewed

	@@ -0,0 +1,292 @@

+# Absent standard TAR IDs within 0..highest observed ID, from the pinned source and HF inventory.
+# Source absence does not prove that the corresponding sample data exist elsewhere.
+data/train/interact-env/000470.tar
+data/train/interact-env/000471.tar
+data/train/interact-env/000472.tar
+data/train/interact-env/000473.tar
+data/train/interact-env/000474.tar
+data/train/interact-env/000495.tar
+data/train/interact-env/000496.tar
+data/train/interact-env/000497.tar
+data/train/interact-env/000498.tar
+data/train/interact-env/000499.tar
+data/train/interact-env/000520.tar
+data/train/interact-env/000521.tar
+data/train/interact-env/000522.tar
+data/train/interact-env/000523.tar
+data/train/interact-env/000524.tar
+data/train/interact-env/000545.tar
+data/train/interact-env/000546.tar
+data/train/interact-env/000547.tar
+data/train/interact-env/000548.tar
+data/train/interact-env/000549.tar
+data/train/interact-env/000570.tar
+data/train/interact-env/000571.tar
+data/train/interact-env/000572.tar
+data/train/interact-env/000573.tar
+data/train/interact-env/000574.tar
+data/train/interact-env/000595.tar
+data/train/interact-env/000596.tar
+data/train/interact-env/000597.tar
+data/train/interact-env/000598.tar
+data/train/interact-env/000599.tar
+data/train/interact-env/000620.tar
+data/train/interact-env/000621.tar
+data/train/interact-env/000622.tar
+data/train/interact-env/000623.tar
+data/train/interact-env/000624.tar
+data/train/interact-env/000645.tar
+data/train/interact-env/000646.tar
+data/train/interact-env/000647.tar
+data/train/interact-env/000648.tar
+data/train/interact-env/000649.tar
+data/train/interact-env/000670.tar
+data/train/interact-env/000671.tar
+data/train/interact-env/000672.tar
+data/train/interact-env/000673.tar
+data/train/interact-env/000674.tar
+data/train/interact-env/000695.tar
+data/train/interact-env/000696.tar
+data/train/interact-env/000697.tar
+data/train/interact-env/000698.tar
+data/train/interact-env/000699.tar
+data/train/interact-env/000720.tar
+data/train/interact-env/000721.tar
+data/train/interact-env/000722.tar
+data/train/interact-env/000723.tar
+data/train/interact-env/000724.tar
+data/train/interact-env/000745.tar
+data/train/interact-env/000746.tar
+data/train/interact-env/000747.tar
+data/train/interact-env/000748.tar
+data/train/interact-env/000749.tar
+data/train/interact-env/000770.tar
+data/train/interact-env/000771.tar
+data/train/interact-env/000772.tar
+data/train/interact-env/000773.tar
+data/train/interact-env/000774.tar
+data/train/interact-env/000795.tar
+data/train/interact-env/000796.tar
+data/train/interact-env/000797.tar
+data/train/interact-env/000798.tar
+data/train/interact-env/000799.tar
+data/train/interact-env/000820.tar
+data/train/interact-env/000821.tar
+data/train/interact-env/000822.tar
+data/train/interact-env/000823.tar
+data/train/interact-env/000824.tar
+data/train/interact-env/000845.tar
+data/train/interact-env/000846.tar
+data/train/interact-env/000847.tar
+data/train/interact-env/000848.tar
+data/train/interact-env/000849.tar
+data/train/interact-env/000870.tar
+data/train/interact-env/000871.tar
+data/train/interact-env/000872.tar
+data/train/interact-env/000873.tar
+data/train/interact-env/000874.tar
+data/train/interact-env/000895.tar
+data/train/interact-env/000896.tar
+data/train/interact-env/000897.tar
+data/train/interact-env/000898.tar
+data/train/interact-env/000899.tar
+data/train/interact-env/000920.tar
+data/train/interact-env/000921.tar
+data/train/interact-env/000922.tar
+data/train/interact-env/000923.tar
+data/train/interact-env/000924.tar
+data/train/interact-env/000945.tar
+data/train/interact-env/000946.tar
+data/train/interact-env/000947.tar
+data/train/interact-env/000948.tar
+data/train/interact-env/000949.tar
+data/train/interact-env/000970.tar
+data/train/interact-env/000971.tar
+data/train/interact-env/000972.tar
+data/train/interact-env/000973.tar
+data/train/interact-env/000974.tar
+data/train/interact-env/000995.tar
+data/train/interact-env/000996.tar
+data/train/interact-env/000997.tar
+data/train/interact-env/000998.tar
+data/train/interact-env/000999.tar
+data/train/interact-env/001020.tar
+data/train/interact-env/001021.tar
+data/train/interact-env/001022.tar
+data/train/interact-env/001023.tar
+data/train/interact-env/001024.tar
+data/train/interact-env/001045.tar
+data/train/interact-env/001046.tar
+data/train/interact-env/001047.tar
+data/train/interact-env/001048.tar
+data/train/interact-env/001049.tar
+data/train/interact-env/001070.tar
+data/train/interact-env/001071.tar
+data/train/interact-env/001072.tar
+data/train/interact-env/001073.tar
+data/train/interact-env/001074.tar
+data/train/interact-env/001095.tar
+data/train/interact-env/001096.tar
+data/train/interact-env/001097.tar
+data/train/interact-env/001098.tar
+data/train/interact-env/001099.tar
+data/train/interact-env/001120.tar
+data/train/interact-env/001121.tar
+data/train/interact-env/001122.tar
+data/train/interact-env/001123.tar
+data/train/interact-env/001124.tar
+data/train/interact-env/001145.tar
+data/train/interact-env/001146.tar
+data/train/interact-env/001147.tar
+data/train/interact-env/001148.tar
+data/train/interact-env/001149.tar
+data/train/interact-env/001170.tar
+data/train/interact-env/001171.tar
+data/train/interact-env/001172.tar
+data/train/interact-env/001173.tar
+data/train/interact-env/001174.tar
+data/train/interact-env/001195.tar
+data/train/interact-env/001196.tar
+data/train/interact-env/001197.tar
+data/train/interact-env/001198.tar
+data/train/interact-env/001199.tar
+data/train/interact-env/001220.tar
+data/train/interact-env/001221.tar
+data/train/interact-env/001222.tar
+data/train/interact-env/001223.tar
+data/train/interact-env/001224.tar
+data/train/interact-env/001245.tar
+data/train/interact-env/001246.tar
+data/train/interact-env/001247.tar
+data/train/interact-env/001248.tar
+data/train/interact-env/001249.tar
+data/train/interact-env/001270.tar
+data/train/interact-env/001271.tar
+data/train/interact-env/001272.tar
+data/train/interact-env/001273.tar
+data/train/interact-env/001274.tar
+data/train/interact-env/001295.tar
+data/train/interact-env/001296.tar
+data/train/interact-env/001297.tar
+data/train/interact-env/001298.tar
+data/train/interact-env/001299.tar
+data/train/interact-env/001320.tar
+data/train/interact-env/001321.tar
+data/train/interact-env/001322.tar
+data/train/interact-env/001323.tar
+data/train/interact-env/001324.tar
+data/train/interact-env/001345.tar
+data/train/interact-env/001346.tar
+data/train/interact-env/001347.tar
+data/train/interact-env/001348.tar
+data/train/interact-env/001349.tar
+data/train/interact-env/001370.tar
+data/train/interact-env/001371.tar
+data/train/interact-env/001372.tar
+data/train/interact-env/001373.tar
+data/train/interact-env/001374.tar
+data/train/interact-env/001395.tar
+data/train/interact-env/001396.tar
+data/train/interact-env/001397.tar
+data/train/interact-env/001398.tar
+data/train/interact-env/001399.tar
+data/train/interact-env/001420.tar
+data/train/interact-env/001421.tar
+data/train/interact-env/001422.tar
+data/train/interact-env/001423.tar
+data/train/interact-env/001424.tar
+data/train/interact-env/001445.tar
+data/train/interact-env/001446.tar
+data/train/interact-env/001447.tar
+data/train/interact-env/001448.tar
+data/train/interact-env/001449.tar
+data/train/interact-env/001470.tar
+data/train/interact-env/001471.tar
+data/train/interact-env/001472.tar
+data/train/interact-env/001473.tar
+data/train/interact-env/001474.tar
+data/train/interact-env/001495.tar
+data/train/interact-env/001496.tar
+data/train/interact-env/001497.tar
+data/train/interact-env/001498.tar
+data/train/interact-env/001499.tar
+data/train/interact-env/001520.tar
+data/train/interact-env/001521.tar
+data/train/interact-env/001522.tar
+data/train/interact-env/001523.tar
+data/train/interact-env/001524.tar
+data/train/interact-env/001545.tar
+data/train/interact-env/001546.tar
+data/train/interact-env/001547.tar
+data/train/interact-env/001548.tar
+data/train/interact-env/001549.tar
+data/train/interact-env/001570.tar
+data/train/interact-env/001571.tar
+data/train/interact-env/001572.tar
+data/train/interact-env/001573.tar
+data/train/interact-env/001574.tar
+data/train/interact-env/001595.tar
+data/train/interact-env/001596.tar
+data/train/interact-env/001597.tar
+data/train/interact-env/001598.tar
+data/train/interact-env/001599.tar
+data/train/interact-env/001620.tar
+data/train/interact-env/001621.tar
+data/train/interact-env/001622.tar
+data/train/interact-env/001623.tar
+data/train/interact-env/001624.tar
+data/train/interact-env/001645.tar
+data/train/interact-env/001646.tar
+data/train/interact-env/001647.tar
+data/train/interact-env/001648.tar
+data/train/interact-env/001649.tar
+data/train/interact-env/001670.tar
+data/train/interact-env/001671.tar
+data/train/interact-env/001672.tar
+data/train/interact-env/001673.tar
+data/train/interact-env/001674.tar
+data/train/interact-env/001695.tar
+data/train/interact-env/001696.tar
+data/train/interact-env/001697.tar
+data/train/interact-env/001698.tar
+data/train/interact-env/001699.tar
+data/train/interact-env/001720.tar
+data/train/interact-env/001721.tar
+data/train/interact-env/001722.tar
+data/train/interact-env/001723.tar
+data/train/interact-env/001724.tar
+data/train/interact-env/001745.tar
+data/train/interact-env/001746.tar
+data/train/interact-env/001747.tar
+data/train/interact-env/001748.tar
+data/train/interact-env/001749.tar
+data/train/interact-env/001770.tar
+data/train/interact-env/001771.tar
+data/train/interact-env/001772.tar
+data/train/interact-env/001773.tar
+data/train/interact-env/001774.tar
+data/train/interact-env/001795.tar
+data/train/interact-env/001796.tar
+data/train/interact-env/001797.tar
+data/train/interact-env/001798.tar
+data/train/interact-env/001799.tar
+data/train/interact-env/001820.tar
+data/train/interact-env/001821.tar
+data/train/interact-env/001822.tar
+data/train/interact-env/001823.tar
+data/train/interact-env/001824.tar
+data/train/interact-env/001845.tar
+data/train/interact-env/001846.tar
+data/train/interact-env/001847.tar
+data/train/interact-env/001848.tar
+data/train/interact-env/001849.tar
+data/train/interact-env/001870.tar
+data/train/interact-env/001871.tar
+data/train/interact-env/001872.tar
+data/train/interact-env/001873.tar
+data/train/interact-env/001874.tar
+data/train/interact-env/001895.tar
+data/train/interact-env/001896.tar
+data/train/interact-env/001897.tar
+data/train/interact-env/001898.tar
+data/train/interact-env/001899.tar

release_audit.json ADDED Viewed

	@@ -0,0 +1,582 @@

+{
+  "audited_at_utc": "2026-10-13T15:31:14.432064+00:00",
+  "method": "HF LFS size/SHA-256 metadata at one immutable revision; no TAR payloads downloaded or decoded",
+  "numbering_scope": "Per folder, numeric filename suffix, from zero to highest observed ID",
+  "manifest_tar_count": 6157,
+  "remote_tar_count": 6157,
+  "samples": 6304768,
+  "total_size_bytes": 764162375680,
+  "missing_remote_paths": [],
+  "unexpected_remote_paths": [],
+  "size_or_sha256_mismatch": [],
+  "invalid_manifest_records": [],
+  "duplicate_manifest_paths": [],
+  "summary_matches_inventory": true,
+  "remote_matches_manifest": true,
+  "numbering_consecutive": false,
+  "numbering_gap_count": 290,
+  "folders": [
+    {
+      "folder": "data/train/interact-env",
+      "tar_count": 1630,
+      "samples": 1669120,
+      "first_id": 0,
+      "last_id": 1919,
+      "missing_id_count": 290,
+      "missing_id_ranges": [
+        [
+          470,
+          474
+        ],
+        [
+          495,
+          499
+        ],
+        [
+          520,
+          524
+        ],
+        [
+          545,
+          549
+        ],
+        [
+          570,
+          574
+        ],
+        [
+          595,
+          599
+        ],
+        [
+          620,
+          624
+        ],
+        [
+          645,
+          649
+        ],
+        [
+          670,
+          674
+        ],
+        [
+          695,
+          699
+        ],
+        [
+          720,
+          724
+        ],
+        [
+          745,
+          749
+        ],
+        [
+          770,
+          774
+        ],
+        [
+          795,
+          799
+        ],
+        [
+          820,
+          824
+        ],
+        [
+          845,
+          849
+        ],
+        [
+          870,
+          874
+        ],
+        [
+          895,
+          899
+        ],
+        [
+          920,
+          924
+        ],
+        [
+          945,
+          949
+        ],
+        [
+          970,
+          974
+        ],
+        [
+          995,
+          999
+        ],
+        [
+          1020,
+          1024
+        ],
+        [
+          1045,
+          1049
+        ],
+        [
+          1070,
+          1074
+        ],
+        [
+          1095,
+          1099
+        ],
+        [
+          1120,
+          1124
+        ],
+        [
+          1145,
+          1149
+        ],
+        [
+          1170,
+          1174
+        ],
+        [
+          1195,
+          1199
+        ],
+        [
+          1220,
+          1224
+        ],
+        [
+          1245,
+          1249
+        ],
+        [
+          1270,
+          1274
+        ],
+        [
+          1295,
+          1299
+        ],
+        [
+          1320,
+          1324
+        ],
+        [
+          1345,
+          1349
+        ],
+        [
+          1370,
+          1374
+        ],
+        [
+          1395,
+          1399
+        ],
+        [
+          1420,
+          1424
+        ],
+        [
+          1445,
+          1449
+        ],
+        [
+          1470,
+          1474
+        ],
+        [
+          1495,
+          1499
+        ],
+        [
+          1520,
+          1524
+        ],
+        [
+          1545,
+          1549
+        ],
+        [
+          1570,
+          1574
+        ],
+        [
+          1595,
+          1599
+        ],
+        [
+          1620,
+          1624
+        ],
+        [
+          1645,
+          1649
+        ],
+        [
+          1670,
+          1674
+        ],
+        [
+          1695,
+          1699
+        ],
+        [
+          1720,
+          1724
+        ],
+        [
+          1745,
+          1749
+        ],
+        [
+          1770,
+          1774
+        ],
+        [
+          1795,
+          1799
+        ],
+        [
+          1820,
+          1824
+        ],
+        [
+          1845,
+          1849
+        ],
+        [
+          1870,
+          1874
+        ],
+        [
+          1895,
+          1899
+        ]
+      ],
+      "duplicate_numeric_ids": [],
+      "filename_prefixes": [
+        {
+          "prefix": "",
+          "count": 1630,
+          "first_id": 0,
+          "last_id": 1919
+        }
+      ]
+    },
+    {
+      "folder": "data/train/interact-indoor",
+      "tar_count": 2087,
+      "samples": 2137088,
+      "first_id": 0,
+      "last_id": 2086,
+      "missing_id_count": 0,
+      "missing_id_ranges": [],
+      "duplicate_numeric_ids": [],
+      "filename_prefixes": [
+        {
+          "prefix": "",
+          "count": 1090,
+          "first_id": 0,
+          "last_id": 1089
+        },
+        {
+          "prefix": "indoor_",
+          "count": 997,
+          "first_id": 1090,
+          "last_id": 2086
+        }
+      ]
+    },
+    {
+      "folder": "data/train/single",
+      "tar_count": 2440,
+      "samples": 2498560,
+      "first_id": 0,
+      "last_id": 2439,
+      "missing_id_count": 0,
+      "missing_id_ranges": [],
+      "duplicate_numeric_ids": [],
+      "filename_prefixes": [
+        {
+          "prefix": "",
+          "count": 2440,
+          "first_id": 0,
+          "last_id": 2439
+        }
+      ]
+    }
+  ],
+  "pinned_source": {
+    "persistent_id": "doi:10.7910/DVN/VBYEK3",
+    "version": ":draft",
+    "catalogue_bundle_count": 997,
+    "represented_bundle_count": 997,
+    "missing_bundle_ids": [],
+    "unexpected_bundle_ids": [],
+    "bundle_range_issues": [],
+    "unparsed_bundle_ids": [],
+    "numbering": [
+      {
+        "directory": "train/dpx",
+        "present_id_count": 2440,
+        "last_id": 2439,
+        "missing_id_ranges": []
+      },
+      {
+        "directory": "train/graspxl_env",
+        "present_id_count": 1630,
+        "last_id": 1919,
+        "missing_id_ranges": [
+          [
+            470,
+            474
+          ],
+          [
+            495,
+            499
+          ],
+          [
+            520,
+            524
+          ],
+          [
+            545,
+            549
+          ],
+          [
+            570,
+            574
+          ],
+          [
+            595,
+            599
+          ],
+          [
+            620,
+            624
+          ],
+          [
+            645,
+            649
+          ],
+          [
+            670,
+            674
+          ],
+          [
+            695,
+            699
+          ],
+          [
+            720,
+            724
+          ],
+          [
+            745,
+            749
+          ],
+          [
+            770,
+            774
+          ],
+          [
+            795,
+            799
+          ],
+          [
+            820,
+            824
+          ],
+          [
+            845,
+            849
+          ],
+          [
+            870,
+            874
+          ],
+          [
+            895,
+            899
+          ],
+          [
+            920,
+            924
+          ],
+          [
+            945,
+            949
+          ],
+          [
+            970,
+            974
+          ],
+          [
+            995,
+            999
+          ],
+          [
+            1020,
+            1024
+          ],
+          [
+            1045,
+            1049
+          ],
+          [
+            1070,
+            1074
+          ],
+          [
+            1095,
+            1099
+          ],
+          [
+            1120,
+            1124
+          ],
+          [
+            1145,
+            1149
+          ],
+          [
+            1170,
+            1174
+          ],
+          [
+            1195,
+            1199
+          ],
+          [
+            1220,
+            1224
+          ],
+          [
+            1245,
+            1249
+          ],
+          [
+            1270,
+            1274
+          ],
+          [
+            1295,
+            1299
+          ],
+          [
+            1320,
+            1324
+          ],
+          [
+            1345,
+            1349
+          ],
+          [
+            1370,
+            1374
+          ],
+          [
+            1395,
+            1399
+          ],
+          [
+            1420,
+            1424
+          ],
+          [
+            1445,
+            1449
+          ],
+          [
+            1470,
+            1474
+          ],
+          [
+            1495,
+            1499
+          ],
+          [
+            1520,
+            1524
+          ],
+          [
+            1545,
+            1549
+          ],
+          [
+            1570,
+            1574
+          ],
+          [
+            1595,
+            1599
+          ],
+          [
+            1620,
+            1624
+          ],
+          [
+            1645,
+            1649
+          ],
+          [
+            1670,
+            1674
+          ],
+          [
+            1695,
+            1699
+          ],
+          [
+            1720,
+            1724
+          ],
+          [
+            1745,
+            1749
+          ],
+          [
+            1770,
+            1774
+          ],
+          [
+            1795,
+            1799
+          ],
+          [
+            1820,
+            1824
+          ],
+          [
+            1845,
+            1849
+          ],
+          [
+            1870,
+            1874
+          ],
+          [
+            1895,
+            1899
+          ]
+        ]
+      },
+      {
+        "directory": "train/graspxl_indoor",
+        "present_id_count": 2087,
+        "last_id": 2086,
+        "missing_id_ranges": []
+      }
+    ],
+    "all_catalogued_ranges_present": true
+  },
+  "caveat": "Source completeness refers only to the supplied pinned catalogue. Numbering gaps do not prove missing content beyond that catalogue; sample counts come from the previously validated manifest.",
+  "repo_id": "chen-si-02/AnyHand-Dataset",
+  "revision": "cc64086d7c3d99ccbfbb2e4b191726541eefa74a",
+  "source_catalogue_sha256": "43bf86d8809d6a92a822da9cb77bc0764777e7c69711a920642228682ffde39a",
+  "renamed_tar_count": 3717
+}

transfer_summary.json CHANGED Viewed

@@ -1,12 +1,16 @@
-{
-  "shards": {
-    "uploaded": 6157
-  },
-  "bundles": {
-    "done": 997
-  },
-  "uploaded_samples": 6304768,
-  "pipeline_finished": true,
-  "remote_audit_commit": "0d6c2be3961e55f95ae40773123f44e82aec80c0",
-  "complete": true
-}

+{
+  "shards": {
+    "uploaded": 6157
+  },
+  "bundles": {
+    "done": 997
+  },
+  "uploaded_samples": 6304768,
+  "pipeline_finished": true,
+  "remote_audit_commit": "cc64086d7c3d99ccbfbb2e4b191726541eefa74a",
+  "complete": true,
+  "numbering_consecutive": false,
+  "numbering_gap_count": 290,
+  "completeness_scope": "All bundles and numeric member ranges in the pinned Dataverse catalogue; numbering is not gap-free",
+  "release_audit_file": "release_audit.json"
+}