Université de Namur
Deep Learning Applied to Sign Language
This thesis aims to make a first step towards an algorithm able to translate French Belgian Sign Language to French. It explores existing methods developed in video recognition, in particular deep learning. Transfer learning is used to recognize signs based on frames extracted from video recordings. The Sign Language MNIST dataset is proposed in order to illustrate preliminary results.
Jérôme Fink holds a Master degree in Computer Science with a major in Data Science from the Université de Namur. He received the third price at the HEX2018 hackaton with a project on electric buses fleet optimization. In 2017, he earned the Data Driven Innovation prize for the Citizen of Wallonia hackathon for an application to sell solar electricity.