translate

Deaf and Hard of Hearing

Australian Sign Language signs (High Quality) Data Set

M. Kadous

The studyaims for recognition of Australian Sign Language using instrumented gloves. Samples from a single signer (a native Auslan signer) were collected over a period of nine weeks. In total, 27 samples per sign, and a total of 2565 signs were collected. The average length of each sign was approximately 57 frames.
Tags: 
Sign language

ASL-100-RGBD: An Isolated-Signing RGBD Dataset of 100 American Sign Language Signs Produced by Fluent ASL Signers

S. Hassan, L. Berke, E. Vahdani, L. Jing, Y. Tian, M. Huenerfauth

A RGBD dataset of body movements and HD face data for American sign language is collected to aid in the development of a sign-recognition system. Kinect 2.0 RGBD camera captured the 100 signs of 22 fluent ASL signers, yeidling a total collection of 42 videos, and the recordings were seperated into skeleton, face markings and depth maps and annotated together.
Tags: 
Sign languageAmerican Sign Language

National Center for Sign Language and Gesture Resources (NCSLGR) Corpus

C. Neidle

This project makes available several different types of experimental resources and analyzed data to facilitate linguistic and computational research on signed languages and the gestural components of spoken languages. ASL videos were collected and linguistically annotated.
Tags: 
Sign language

RWTH-PHOENIX-Weather: A Large Vocabulary Sign Language Recognition and Translation Corpus

J. Forster, C. Schmidt, T. Hoyoux, O. Koller, U. Zelle, J. Piater, H. Ney

German sign language transcriptions of weather announcements are collected. 190 weather forecasts of one and a half minutes are recorded and manually annotated using glosses distinguishing sign variants. Time boundaries have been marked on the sentence and the gloss level.
Tags: 
Sign languageGermanGerman sign language

Content4All Open Research Sign Language Translation Datasets

N. Camgoz, B. Saunders, G. Rochette, M. Giovanelli, G. Inches, R. Nachtrab-Ribback, R. Bowden

This work releases six datasets containing 20 hours of broadcast footage annotated by Deaf experts and interpreters for automatic sign language translation research. Raw footage is anonymized and processed using OpenPose to extract 2D and 3D body pose information, and spoken language text to sign language videos are aligned as part of the annotation process.
Tags: 
Sign language

MEDIAPI-SKEL

H. Bull, A. Braffort, M. Gouiffès

A 2D-skeleton database of translations of TV programs into French sign language aligned with French subtitles French Sign Language 2D skeleton videos aligned with French subtitles. Videos capture journalists using sign language.
Tags: 
Sign language

PSL ToF 84

T. Kapuscinski, M. Oszust, M. Wysocki, D. Warchoł

A 3D video, point cloud dataset is collected to explore sign language recognition using skeleton tracking and hand skin recognition. A single participant performed 84 signs 20 times, this was recorded using 3D Kinect sensors and the point cloud data was collected.
Tags: 
Sign languagePolishPolish sign language

RWTH-BOSTON-50

M. Zahedi, D. Keysers, T. Deselaers, H. Ney.

A dataset of video streams of American sign language sentences is collected Multiple cameras collect videos of ASL signers, and then the video is manually annotated. 50 sign language words are signed by 3 signers to generate 201 annotated videos.
Tags: 
Sign languageEnglishAmerican Sign Language

ArASL: Arabic Alphabets Sign Language Dataset

G. Latif, N. Mohammad, J. Alghazo, R. AlKhalaf, R. AlKhalaf

A large fully-labelled dataset for Arabic Sign Language (ArSL) images is collected for research in automated systems using machine learning and computer vision for the deaf and hard of hearing people. A smart camera attached to tripod was used to capture the images and partiicpants were asked to stand around 1 m away from the camera.
Tags: 
Sign languagearabicArabic sign language

PSL Kinect 30

M. Oszust, M. Wysocki

A 3D video, point cloud dataset is collected to explore sign language recognition using skeleton tracking and hand skin recognition. A single participant performed 30 signs 10 times, this was recorded using 3D Kinect sensors and the point cloud data was collected.
Tags: 
Sign languagePolishPolish sign language