File size: 8,042 Bytes
39fd722
 
 
 
99f6cba
 
 
 
 
 
39fd722
 
 
 
99f6cba
39fd722
99f6cba
39fd722
99f6cba
 
 
39fd722
 
99f6cba
 
 
39fd722
99f6cba
39fd722
99f6cba
39fd722
99f6cba
 
 
 
 
 
 
 
39fd722
 
 
99f6cba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39fd722
 
99f6cba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39fd722
99f6cba
39fd722
99f6cba
39fd722
99f6cba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
39fd722
 
 
 
 
99f6cba
39fd722
99f6cba
 
 
 
 
 
39fd722
99f6cba
 
 
 
 
 
 
 
 
39fd722
 
99f6cba
 
 
 
 
 
 
 
 
39fd722
99f6cba
 
 
 
 
39fd722
99f6cba
39fd722
 
 
99f6cba
39fd722
 
 
 
 
31fa9bd
99f6cba
 
 
 
 
39fd722
 
99f6cba
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
---
license: cc-by-nc-sa-4.0
language:
  - ms
  - zsm
pretty_name: MySign  Malaysian Sign Language 3D Motion Capture
size_categories:
  - 1K<n<10K
task_categories:
  - other
tags:
  - sign-language
  - bahasa-isyarat-malaysia
  - bim
  - malaysian-sign-language
  - motion-capture
  - 3d
  - fbx
  - skeletal-animation
  - gloss
configs: []
---

# MySign

A 3D motion-capture dataset of Malaysian Sign Language (Bahasa Isyarat Malaysia, BIM), distributed as Filmbox (`.fbx`) skeletal animation files.

> **Note**: this dataset is **not** loadable via `datasets.load_dataset(...)` as a tabular split, because the data itself is binary `.fbx`. `metadata.csv` is an index over the FBX files; consumers must read the file at `file_name` to access the motion data.

## Dataset Summary

- **Modality**: 3D skeletal motion capture (`.fbx`)
- **Language**: Malaysian Sign Language (BIM) — ISO `ms` / `zsm`
- **Signers**: 5 (`Signer001``Signer005`)
- **Sample**: one `.fbx` per `(signer, gloss, take)`
- **License**: CC BY-NC-SA 4.0
- **Source repository**: <https://huggingface.co/datasets/mysigner/MySign>

## Repository Structure

```
MySign/
├── Signer001/
│   ├── Above.fbx
│   ├── Apologize.fbx
│   ├── Apologize.001.fbx
│   └── ...
├── Signer002/
├── Signer003/
├── Signer004/
├── Signer005/
├── metadata.csv      # index over all .fbx files
├── croissant.json    # Croissant 1.0 ML-dataset metadata (JSON-LD)
├── README.md
└── .gitattributes
```

## `metadata.csv` Schema

`metadata.csv` is a UTF-8 CSV with a header row. **Each row corresponds to exactly one `.fbx` file in the repository.**

| Column      | Type    | Description |
|-------------|---------|-------------|
| `file_name` | string  | Repository-relative path to the `.fbx`, e.g. `Signer001/Above.fbx`. |
| `gloss`     | string  | Normalized gloss label (UPPERCASE). See [Gloss Normalization](#gloss-normalization). |
| `signer_id` | string  | One of `Signer001``Signer005`. |
| `take`      | integer | Take number for the `(signer_id, gloss)` pair. The original recording is `1`; Blender-style duplicate suffixes `.001`, `.002` map to takes `2`, `3`. |

### Example rows

```csv
file_name,gloss,signer_id,take
Signer001/Above.fbx,ABOVE,Signer001,1
Signer001/Apologize.fbx,APOLOGIZE,Signer001,1
Signer001/Apologize.001.fbx,APOLOGIZE,Signer001,2
Signer002/Actor,_Actress.fbx,ACTOR / ACTRESS,Signer002,1
Signer003/1-hr.fbx,1 HOUR,Signer003,1
Signer004/Less (I).fbx,LESS(I),Signer004,1
```

> **Path quoting**: some `file_name` values contain commas (e.g. `Signer002/Actor,_Actress.fbx`). The CSV is written with Python's default `csv.writer`, which quotes such fields automatically. Read it with any standard CSV parser (pandas, `csv.DictReader`).

### Gloss Normalization

The `gloss` column is derived from each filename and normalized so that the same sign performed by different signers gets the same label. The normalization, applied by `generate_metadata_remote.py`, is:

1. Strip Blender suffix `.001` / `.002` (recorded into `take`).
2. Replace `_` and `-` with spaces.
3. Treat `,` and `;` as synonym separators, joined as `WORD_A / WORD_B`.
4. Uppercase.
5. Expand time-unit abbreviations: `HR→HOUR`, `MIN→MINUTE`, `MTH→MONTH`, `WK→WEEK`, `YR→YEAR`, `SEC→SECOND` (and plurals).
6. Normalize whitespace around parentheses: `WORD (X)``WORD(X)`.
7. Strip unbalanced trailing `)`.
8. Plural→singular merge: when the corpus contains both forms of a word (e.g. `COURSES` and `COURSE`), the plural is rewritten to the singular. A small block-list keeps semantically distinct plurals (`NEWS`, `SHORTS`, `MATHEMATICS`) as plurals.
9. A small hand-curated override map (currently: `COCHLEAR / IMPLANT``COCHLEAR IMPLANT`) fixes cases that no general rule can fix safely.

A full audit of the plural→singular merges is printed by the script every time it runs; check that list before publishing.

## `.fbx` Files

- Format: Autodesk Filmbox (binary `.fbx`).
- One file per take.
- The `file_name` column of `metadata.csv` is the canonical pointer to each file. The same path can be resolved as `https://huggingface.co/datasets/mysigner/MySign/resolve/main/<file_name>`.

## Croissant Metadata

A Croissant 1.0 (JSON-LD) description is provided as `croissant.json`. It declares:

- The repository as a `cr:FileObject` (`encodingFormat: git+https`).
- `metadata.csv` as a `cr:FileObject` of type `text/csv`.
- All `Signer*/*.fbx` files as a `cr:FileSet` named `fbx-files` (`encodingFormat: model/vnd.fbx`).
- A `cr:RecordSet` named `signs` exposing the four columns above. The `file_name` field carries `references: fbx-files`, which tells Croissant consumers the value is a path into the FBX FileSet.

**Limitation**: Croissant 1.0 has no native semantics for FBX. `mlcroissant` will treat every FBX as an opaque binary resource — it will not parse skeletons, animation curves, or any FBX-internal structure. Consumers must use a real FBX library (e.g. Autodesk FBX SDK, Blender, [`pyfbx`](https://pypi.org/project/pyfbx/)) to read the motion data.

## Usage

### List records via `metadata.csv`

```python
import pandas as pd

df = pd.read_csv(
    "https://huggingface.co/datasets/mysigner/MySign/resolve/main/metadata.csv"
)
print(df.head())
print(df["signer_id"].value_counts().sort_index())
print(df["gloss"].nunique(), "unique glosses")
```

### Download a specific FBX

```python
from huggingface_hub import hf_hub_download

local_path = hf_hub_download(
    repo_id="mysigner/MySign",
    filename="Signer001/Above.fbx",
    repo_type="dataset",
)
print(local_path)  # local cache path; open with your FBX library
```

### Inspect via Croissant (mlcroissant)

```python
import mlcroissant as mlc

ds = mlc.Dataset(
    "https://huggingface.co/datasets/mysigner/MySign/resolve/main/croissant.json"
)
print(ds.metadata.name, "—", ds.metadata.description[:80])

for i, rec in enumerate(ds.records(record_set="signs")):
    print(rec)
    if i >= 4:
        break
```

`ds.records("signs")` yields one Python dict per `.fbx`, with `file_name`, `gloss`, `signer_id`, `take`. The FBX bytes are not auto-loaded — read them yourself from `file_name`.

## License

This dataset is released under [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/). You may share and adapt the work for non-commercial purposes, provided you give appropriate credit and distribute derivative works under the same license.

## Citation

```bibtex
@misc{mysign2026,
  title        = {MySign: A High-Fidelity Motion-Capture Dataset for 3D Sign Generation in Bahasa Isyarat Malaysia},
  author       = {{mysigner}},
  year         = {2026},
  howpublished = {Hugging Face Datasets},
  url          = {https://huggingface.co/datasets/mysigner/MySign},
  note         = {CC BY-NC-SA 4.0}
}
```

(Replace with the canonical citation when a paper or technical report becomes available.)

## Limitations

- **Not auto-loadable**: Hugging Face's Dataset Viewer cannot render `.fbx`. The Viewer will not work for this dataset, and `datasets.load_dataset("mysigner/MySign")` will not produce a usable split. Use the workflows above instead.
- **Gloss is filename-derived**: glosses come from filenames written by signers/annotators with slightly different conventions, then normalized. Some collisions may still exist; run the script's plural-merge audit before relying on a particular gloss inventory.
- **Five signers**: signer-conditioned models trained on MySign will have limited speaker coverage.
- **Take semantics are best-effort**: the `take` column distinguishes Blender-duplicated files (`.001`, `.002`) from the original recording, but it does not encode whether a take was a clean recording or a retry.
- **No segmentation, no transcription**: each FBX is one isolated sign. The dataset does not include continuous-signing video, glossed sentences, or non-manual annotations.