Task Mate Kenyan Sign Language Classification Challenge

Being able to communicate through spoken language is one of the things that many of us take for granted. As we learn about each other’s languages and/or specialise in the current lingua franca, we aim to cross the boundaries that our languages inherently create in between them. With this in mind, it is often assumed that sign language can be the alternative for a global way of communication and all the deaf communities around the world can easily communicate with each other. However, the reality is far from the assumption, as there are as many as 300 languages(approximately 50 of them are used in Africa) with even regional dialects within these languages (such as the Black American Sign Language). Although the concept of sign language has been used since 2000 years ago, there’s no functional use of an international standard with various attempts made. This competition is specifically about the Kenyan Sign Language(KSL), which is estimated to be used by more than half of the 600,000 deaf population in Kenya.

The Alphabet of Kenyan Sign Language.

The goal of the competition is to build a model that would be able to identify ten of the common KSL signs used in the language from images. This would help to allow the process of transcribing signed language to written and/or spoken language and make it more accessible. The host company Task Mate also aims to reduce the inequality between the deaf community and the rest of society with this competition to be aligned with the UN's Sustainable Development Goals. An interesting fact about the given dataset is that almost all of the hands shown in the images belong to people of colour as an attempt to minimise the bias in sign language datasets, which are mostly of hands belonging to white population. Our competition team will aim to extract the information from the images through various methods and train their algorithms accordingly to recognize the differences between signs to properly classify them. Follow our social media for further posts about how they will approach the problem!



Previous
Previous

Multi Agent Behavior Challenge

Next
Next

Bloods.AI - Blood Spectroscopy Classification Challenge