Published July 3, 2025 | Version v1.0
Dataset Open

LSE-FS-UVigo Dataset

Description

LSE-FS-UVigo is a dataset of continuous fingerspelling in Spanish Sign Language (LSE), containing 3,010 sequences from 42 signers, recorded under both studio and in-the-wild conditions. The dataset provides frame-level 3D hand keypoints (raw and preprocessed) extracted using MediaPipe, with train/validation/test splits, signer metadata, and, for the subset with RGB distribution permissions (from the LSE-Health-UVigo subset), annotated YouTube timestamps. This resource is intended to advance research in sign language recognition and benchmarking of keypoint-based recognition methods.

Files

README.md

Files (14.7 GB)

Name Size Download all
md5:dd24175f99cbf02fab071faf5047f227
67.9 kB Preview Download
md5:80a48eb608fde276eb147987fea1488c
7.4 GB Preview Download
md5:f7b5d626318d86057388686e37327796
7.3 GB Preview Download
md5:43672886a064348804fa152e09433f2c
1.8 kB Preview Download
md5:a33fe8ab477e46de704d88741e86e225
525 Bytes Preview Download

Additional details

Related works

Is described by
Publication: 10.1109/ACCESS.2025.3631400 (DOI)
Is supplemented by
Dataset: 10.5281/ZENODO.10234464 (DOI)

Funding

Ministerio de Ciencia, Innovación y Universidades
TECNOLOGIAS PARA LA INCLUSION EN LENGUA DE SIGNOS: BASES DE DATOS, RECONOCIMIENTO Y TRADUCCION PID2021-123988OB-C32
Xunta de Galicia
ED431B 2024/36
Xunta de Galicia
Consellería de Educación, Ciencia, Universidades e Formación Profesional and co-funded by UE"

Dates

Submitted
2025-07-03

Software

Repository URL
https://github.com/lruanova/fingerspelling-trainer
Programming language
Python
Development Status
Active