This repository contains the data and code of the paper entitled Automatic recognition of Mexican Sign Language using a depth camera and recurrent neural networks submitted to the journal MDPI Applied Sciences.
The dataset is composed of 3000 samples from 30 different signs. Each sample represents a sign consisting of 20 consecutive frames containing hands, body, and facial 3D keypoint coordinates.
You can download the data here.
The pre-processing, training and evaluation code will be published soon.