American Sign Language Image Dataset (figerspelling original + keypoints + evaluation)

Published: 30 January 2026| Version 1 | DOI: 10.17632/97yd3b5f5h.1
Contributor:
Elisa Cabana Garceran del Vall

Description

The dataset consists of four folders: (1) Dataset Original, (2) Dataset Keypoints, (3) Dataset Black Background, and (4) Controlled Evaluation Dataset. Each folder contains image data, dimension 192 x 192, for each alphabet letter (static only) in American Sign Language (ASL). Under CC BY 4.0 license, researchers can re-use this dataset for ASL fingerspelling recognition and translation, as long as you give appropriate credit by citing the dataset.

Files

Steps to reproduce

All of the details about this dataset and the source code to train models based on this dataset, as well as the final application interface, can be found in the following Github repository: https://github.com/ecabestadistica/ASL-App

Categories

Computer Vision, Image Classification, American Sign Language, Convolutional Neural Network

Licence