How2Sign: a large-scale multimodal dataset for continuous American Sign Language

How2Sign consists of a parallel corpus of 80 hours of sign language videos (collected with multi-view RGB and depth sensor data) with corresponding speech transcriptions and gloss annotations. In addition, a three-hour subset was further recorded in a geodesic dome setup using hundreds of cameras an...

ver descrição completa

Detalhes bibliográficos
Autores: Cardoso Duarte, Amanda, Giró Nieto, Xavier, Palaskar, Shruti, Ghadiyaram, Deepti, Haan, Kenneth de, Metze, Florian, Torres Viñals, Jordi
Formato: conjunto de datos
Fecha de publicación:2024
País:España
Recursos:Consorci de Serveis Universitaris de Catalunya (CSUC)
Repositorio:CORA.Repositori de Dades de Recerca
OAI Identifier:oai:dnet:cora.rdr____::bbca15bbcd803bb46705b03c179c61ed
Acesso em linha:https://doi.org/10.34810/DATA33
Access Level:acceso abierto
Palavra-chave:Computer and Information Science
American Sign Language
Machine translation
Computer vision
Natural language processing
Sign language
Gesture recognition
Descrição
Resumo:How2Sign consists of a parallel corpus of 80 hours of sign language videos (collected with multi-view RGB and depth sensor data) with corresponding speech transcriptions and gloss annotations. In addition, a three-hour subset was further recorded in a geodesic dome setup using hundreds of cameras and sensors, which enables detailed 3D reconstruction and pose estimation and paves the way for vision systems to understand the 3D geometry of sign language.