TReLSU-HS : A new handshape dataset for Uruguayan Sign Language Recognition.

Abstract

In this work we present TReLSU-HS, a new database composed of more than 3000 still images for handshape recognition in the context of Uruguayan Sign Language. TReLSU-HS has 30 classes sampled from 5 native signers. The images were obtained from a previous dataset of Uruguayan Sign Language called Léxico TReLSU. Each component image was labeled according to consistent criteria. This database is useful for the computer science community, especially for designing new sign language recognition methods or to better understand the generalization capability of a given recognition system when it is applied to Uruguayan Sign Language data.