SignLanguage-Dataset-Hub
๐ Sign Language Dataset Hub
[
]
[
]
[
]
A curated, verified catalog of 73+ sign language datasets covering 26 sign languages โ an open collection for sign language recognition (SLR) research, gesture recognition, deaf community technology, and assistive AI development.
๐ฏ Mission
To democratize access to sign language technology by providing:
- Verified dataset links โ every URL checked, sample counts from original sources
- Working tools for loading, visualizing, and processing data
- Demo datasets for learning and prototyping
- Proper attribution to all data creators
Helping developers, researchers, and the deaf community build better assistive technology.
๐ Stats
| Metric | Count |
|---|---|
| Verified Datasets | 73 |
| Sign Languages | 26 |
| Modalities | Video, Image, Sensor, Pose, RGB-D, Skeleton, Text |
| Source Verification | 100% (all URLs checked) |
Datasets by Language
| Language | Code | Datasets | Notable |
|---|---|---|---|
| American Sign Language | ASL | 11 | MS-ASL, WLASL, How2Sign, OpenASL, ASLLVD |
| Arabic Sign Language | ArSL | 2 | ArSL2018, KArSL |
| Australian Sign Language | Auslan | 1 | Auslan Signbank |
| Bangla Sign Language | BdSL | 4 | BdSL47, Ban-Sign-Sent-9K |
| Brazilian Sign Language | Libras | 2 | Libras-UFPR, PHOENIX-Libras |
| British Sign Language | BSL | 3 | BOBSL, BSL Corpus, BSL SignBank |
| Chinese Sign Language | CSL | 2 | DEVISIGN, USTC-CSL |
| Dutch Sign Language | NGT | 1 | CNGT Corpus |
| French Sign Language | LSF | 2 | Dicta-Sign LSF, LSF-Dict |
| German Sign Language | DGS | 3 | RWTH-PHOENIX-2014, PHOENIX-2014T, DGS Corpus |
| Greek Sign Language | GSL | 1 | GSL-50 |
| Indian Sign Language | ISL | 3 | INCLUDE, ISL-CSLTR, ISL-Alphabet |
| Irish Sign Language | ISL | 1 | ISL Corpus |
| Italian Sign Language | LIS | 1 | ATIS |
| Japanese Sign Language | JSL | 1 | J-ASL |
| Korean Sign Language | KSL | 1 | KETI |
| Malaysian Sign Language | BIM | 1 | MSL Dataset |
| Mexican Sign Language | LSM | 1 | LSM Sign Language |
| Russian Sign Language | RSL | 2 | RuSLAN, RSL-Signs |
| Swedish Sign Language | SSL | 1 | SSL Corpus |
| Thai Sign Language | TSL | 1 | TSL-51 |
| Turkish Sign Language | TฤฐD | 1 | AUTSL |
| Multilingual | โ | 5 | SIGN-Hub, Dicta-Sign, SpreadTheSign, OpenSLR, SLP Toolkit |
| Linguistic DBs | โ | 6 | ASL-LEX, BSL SignBank, Auslan Signbank, etc. |
Datasets by Modality
| Modality | Count |
|---|---|
| Video | 35+ |
| Image | 10+ |
| Video + RGB-D + Skeleton | 3 |
| Sensor (IMU/Flex) | 1 |
| Linguistic / Dictionary | 6+ |
| Multilingual Corpus | 5+ |
โ Verification Policy
All dataset source URLs in this repo have been verified. This means:
- โ Every URL resolves (HTTP 200 or auth-gated page)
- โ Sample counts are from the original publication, not estimated
- โ Every dataset has a citation or credit to its creator
- โ No placeholder or fabricated links exist in the active catalog
- โ Datasets we couldnโt verify are excluded (not listed with fake info)
Found a broken link? Please open an issue.
๐ Browse Datasets
See DATASETS.md for the complete verified catalog with:
- 67 datasets across 22 sign languages
- Source URLs, sample counts, licenses, and citations for every entry
- Organized alphabetically by language
๐ Literature & Benchmarks
- REFERENCES.md โ Key papers on sign language recognition, translation, datasets, and pose estimation
- docs/BENCHMARKS.md โ Published accuracy numbers from real papers (WER, BLEU, accuracy)
๐ Quick Start
Clone & Setup
git clone https://github.com/rudra496/SignLanguage-Dataset-Hub.git
cd SignLanguage-Dataset-Hub
pip install -r requirements.txt
Use the Demo Data (Bangla Sign Language Sensor Data)
from scripts.data_loader import BdSLSensorGloveDataset
# Load demo sensor data (4,824 samples)
dataset = BdSLSensorGloveDataset(split='train')
print(f"Loaded {len(dataset)} samples, 36 gesture classes")
sample = dataset[0]
print(f"Gesture: {sample['gesture_name']}")
print(f"Sensors shape: {sample['sensors'].shape}")
Visualize Sensor Data
python tools/visualize.py --data data/bdsl/BdSL-Sensor-Glove/
Browse Programmatically
import pandas as pd
df = pd.read_csv('datasets_catalog.csv')
# Filter by language
asl = df[df['language_code'] == 'ASL']
print(asl[['name', 'samples', 'source_url']])
# Filter by modality
video = df[df['modality'].str.contains('Video')]
print(f"Video datasets: {len(video)}")
Download External Datasets
# From Kaggle (requires Kaggle API key)
pip install kaggle
kaggle datasets download -d datamunge/sign-language-mnist
kaggle datasets download -d grassknoted/asl-alphabet
kaggle datasets download -d ahmedkhan123/arabic-sign-language
# From Hugging Face
pip install datasets
python -c "from datasets import load_dataset; ds = load_dataset('banglagov/Ban-Sign-Sent-9K-V1')"
# From Zenodo
wget https://zenodo.org/record/7067906/files/BdSL47.zip
๐ ๏ธ Included Tools
| Tool | Description | Location |
|---|---|---|
| Data Loader | PyTorch dataset classes for sensor data | scripts/data_loader.py |
| Download Script | Multi-source dataset downloader | scripts/download_datasets.py |
| Visualizer | Sensor data visualization | tools/visualize.py |
| Data Generator | Demo data creation utilities | tools/generate_realistic_data.py |
๐ Repository Structure
SignLanguage-Dataset-Hub/
โโโ DATASETS.md # Complete verified dataset catalog (67 datasets)
โโโ datasets_catalog.csv # Machine-readable catalog
โโโ STATISTICS.md # Detailed statistics & breakdowns
โโโ README.md # This file
โโโ CHANGELOG.md # Version history
โโโ data/
โ โโโ bdsl/
โ โโโ BdSL-Sensor-Glove/ # Demo sensor dataset (4,824 samples)
โโโ docs/
โ โโโ BENCHMARKS.md # Published accuracy numbers & comparisons
โ โโโ LICENSE_ATTRIBUTION.md # Per-dataset license & citation info
โ โโโ TUTORIALS.md # 9 tutorials (beginner to advanced)
โ โโโ QUICKSTART.md # Quick start guide
โ โโโ CONTRIBUTING.md # How to contribute
โโโ scripts/
โ โโโ data_loader.py # PyTorch data loaders
โ โโโ download_datasets.py # Multi-source downloader
โโโ tools/
โ โโโ visualize.py # Sensor data visualization
โ โโโ generate_realistic_data.py # Data generation
โโโ .github/ # Issue templates & PR template
โโโ CITATION.cff # Citation metadata
โโโ LICENSE # CC BY 4.0
โโโ requirements.txt # Python dependencies
๐ Tutorials
We include 9 tutorials from beginner to advanced:
| # | Tutorial | Level |
|---|---|---|
| 1 | Introduction to Sign Language Recognition | Beginner |
| 2 | Loading and Exploring Datasets | Beginner |
| 3 | Visualizing Sign Language Data | Beginner |
| 4 | Building Your First Classifier | Intermediate |
| 5 | Hand Pose Estimation with MediaPipe | Intermediate |
| 6 | Data Augmentation Techniques | Intermediate |
| 7 | Real-time Recognition System | Advanced |
| 8 | Continuous Sign Language Recognition | Advanced |
| 9 | Multilingual Sign Recognition | Advanced |
See docs/TUTORIALS.md and docs/QUICKSTART.md.
๐ Citation
If you use this repository, please cite:
@misc{signlanguage_dataset_hub,
title = {Sign Language Dataset Hub: A Verified Catalog of Sign Language Datasets},
author = {Sarker, Rudra and Contributors},
year = {2026},
url = {https://github.com/rudra496/SignLanguage-Dataset-Hub}
}
Please also cite the original dataset creators when using their data. See docs/LICENSE_ATTRIBUTION.md for per-dataset citation information.
๐ค Contributing
We welcome contributions! See docs/CONTRIBUTING.md.
๐ค Why This Repo?
| Feature | This Repo | Typical SLR Papers/GitHub Lists | Kaggle Collections |
|---|---|---|---|
| Datasets | 73+ curated | 5โ20 mentioned inline | 10โ30, unverified |
| Sign Languages | 26 | 1โ5 | 3โ10 |
| URL Verification | โ All checked | โ Often broken links | โ Mixed |
| Sample Counts | From original sources | Inconsistent | User-reported |
| License Info | โ Per dataset | Rarely included | Rarely included |
| Modality Tags | โ All datasets | Partial | Tags vary |
| Tools & Scripts | โ Included | โ | โ |
| Demo Datasets | โ Included | โ | โ |
| Open Source | CC BY 4.0 | Varies | Varies |
| Actively Maintained | โ | Usually one-time | Community |
Rules:
- Every dataset must have a verifiable source URL
- Sample counts must come from the original source
- Include license and citation information
- No placeholder or fabricated data โ ever
๐ Citation
If you use this dataset catalog in your research, please cite:
@misc{signlanguage_dataset_hub,
title = {Sign Language Dataset Hub: A Curated Catalog of 73+ Verified Datasets for 26 Sign Languages},
author = {Sarker, Rudra},
year = {2025},
url = {https://github.com/rudra496/SignLanguage-Dataset-Hub},
note = {Version 1.0}
}
๐ License
This repository is licensed under CC BY 4.0.
Individual datasets have their own licenses โ see docs/LICENSE_ATTRIBUTION.md for details. Some datasets are research-use only and may require institutional agreements.
๐ Acknowledgments
This hub would not be possible without the researchers and organizations who created and shared these datasets:
- Microsoft Research (MS-ASL)
- Oxford VGG (BOBSL)
- RWTH Aachen University (PHOENIX)
- Boston University (ASLLVD)
- IISc Bangalore (INCLUDE)
- CASIA / USTC (Chinese SL datasets)
- Ankara University (AUTSL)
- University of Hamburg (DGS Corpus)
- UCL DCAL (BSL Corpus, BSL SignBank)
- Stockholm University (SSL Corpus)
- Radboud University (CNGT)
- European Sign Language Centre (SpreadTheSign)
- Macquarie University (Auslan Signbank)
- EU Framework Programme (SIGN-Hub, Dicta-Sign)
- Hugging Face community datasets
- Kaggle community datasets
- And many more individual researchers and Deaf community members
๐ Contact
- Maintainer: Rudra Sarker
- Institution: Shahjalal University of Science and Technology
๐ Connect
Built with โค๏ธ by Rudra Sarker
CC BY 4.0 License ยท Free & Open Source Forever