Titung/nepali-sign-language-video

Nepali Sign Language Video Dataset (NSL23)

First publicly available dataset for Nepali Sign Language (NSL) alphabet recognition.

Dataset Summary

Property	Value
Total videos	630
Total gestures	1,205
Consonants	36
Vowels	13
Volunteers	14 (5 experts, 9 beginners)
Environments	Bright, Dark, Prepared, Unprepared, RealWorld
Format	.mov video files

Splits

Split	Records
train	508
validation	61
test	61

Features

• hf_path — path to video file within this repo (videos/...)
• label — sign label (Devanagari consonant or vowel)
• sign_type — consonant or vowel
• volunteer — signer ID (S1–S14)
• environment — one of Bright / Dark / Prepared / Unprepared / RealWorld
• is_cropped — whether the video is a cropped version
• is_expert — S1–S5 are experts (5–25 years NSL experience)

File Naming Convention

S{volunteer}_NSL_{Consonant|Vowel}_{environment}[_Cropped].mov

Loading

from datasets import load_dataset

# Metadata only
ds = load_dataset("Titung/nepali-sign-language-video")

# Access a record
print(ds["train"][0])

Citation

@article{sunuwar2024nsl23,
  title={NSL23 dataset for alphabets of Nepali sign language},
  author={Sunuwar, Jhuma and Borah, Samarjeet and Kharga, Aditi},
  journal={Data in Brief},
  year={2024},
  publisher={Elsevier},
  doi={10.1016/j.dib.2024.109986}
}