Point at the open data_process pipeline that turns this raw layer into UniML3D
Browse files
README.md
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
language:
|
| 3 |
- en
|
| 4 |
-
license: other
|
| 5 |
size_categories:
|
| 6 |
- 1K<n<10K
|
| 7 |
-
pretty_name: Truebones ZOO Annotations
|
| 8 |
-
license_name: truebones-terms-of-use
|
| 9 |
-
license_link: https://truebones.gumroad.com/l/skZMC
|
| 10 |
tags:
|
| 11 |
- 3d
|
| 12 |
- animation
|
|
@@ -24,14 +24,10 @@ configs:
|
|
| 24 |
data_files: species.csv
|
| 25 |
- config_name: clips
|
| 26 |
data_files: clips.csv
|
| 27 |
-
task_categories:
|
| 28 |
-
- other
|
| 29 |
---
|
| 30 |
|
| 31 |
# Truebones ZOO Annotations
|
| 32 |
|
| 33 |
-
[Project page](https://linzhanmou.com/unimate/) | [Code](https://github.com/Friedrich-M/UniMate)
|
| 34 |
-
|
| 35 |
Text prompts, per-clip metadata, rest-pose renders and the exact build pipeline for
|
| 36 |
**Truebones ZOO** — **1,097 animal motion clips across 74 skeletons**: mammals, birds,
|
| 37 |
reptiles, insects, marine and prehistoric creatures. 1.02 hours, 111,158 frames, uniformly
|
|
@@ -226,6 +222,15 @@ CC-BY-NC-4.0 — non-commercial use only, narrower than the rest of this reposit
|
|
| 226 |
captions are shared by two clips each; retrieval metrics should group ground truth by caption
|
| 227 |
text so those pairs are not scored as misses.
|
| 228 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 229 |
## Usage
|
| 230 |
|
| 231 |
```python
|
|
@@ -292,4 +297,4 @@ This dataset is part of **UniML3D**, the training corpus introduced in
|
|
| 292 |
journal = {arXiv preprint arXiv:2609.05415},
|
| 293 |
year = {2026}
|
| 294 |
}
|
| 295 |
-
```
|
|
|
|
| 1 |
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: truebones-terms-of-use
|
| 4 |
+
license_link: https://truebones.gumroad.com/l/skZMC
|
| 5 |
+
pretty_name: Truebones ZOO Annotations
|
| 6 |
language:
|
| 7 |
- en
|
|
|
|
| 8 |
size_categories:
|
| 9 |
- 1K<n<10K
|
|
|
|
|
|
|
|
|
|
| 10 |
tags:
|
| 11 |
- 3d
|
| 12 |
- animation
|
|
|
|
| 24 |
data_files: species.csv
|
| 25 |
- config_name: clips
|
| 26 |
data_files: clips.csv
|
|
|
|
|
|
|
| 27 |
---
|
| 28 |
|
| 29 |
# Truebones ZOO Annotations
|
| 30 |
|
|
|
|
|
|
|
| 31 |
Text prompts, per-clip metadata, rest-pose renders and the exact build pipeline for
|
| 32 |
**Truebones ZOO** — **1,097 animal motion clips across 74 skeletons**: mammals, birds,
|
| 33 |
reptiles, insects, marine and prehistoric creatures. 1.02 hours, 111,158 frames, uniformly
|
|
|
|
| 222 |
captions are shared by two clips each; retrieval metrics should group ground truth by caption
|
| 223 |
text so those pairs are not scored as misses.
|
| 224 |
|
| 225 |
+
## Processing it into motion data
|
| 226 |
+
|
| 227 |
+
This repository is the **raw** layer. The pipeline that turns it into canonicalized, text-paired motion clips is open at [`UniMate/data_process`](https://github.com/Friedrich-M/UniMate/tree/main/data_process): it exports each clip to NPZ, renders previews, captions them with a vision-language model, cleans the joint labels and derives the training features. The processed release built from it is [UniML3D](https://huggingface.co/datasets/Linzhan/UniML3D).
|
| 228 |
+
|
| 229 |
+
```bash
|
| 230 |
+
bash data_process/scripts/run_download.sh truebones
|
| 231 |
+
bash data_process/scripts/run_export.sh truebones
|
| 232 |
+
```
|
| 233 |
+
|
| 234 |
## Usage
|
| 235 |
|
| 236 |
```python
|
|
|
|
| 297 |
journal = {arXiv preprint arXiv:2609.05415},
|
| 298 |
year = {2026}
|
| 299 |
}
|
| 300 |
+
```
|