Pervasive System's Bachelor student wins top price at international conference challenge on automated handwriting recognition Applying Transformers to digitial character recognition proves to be a winning strategy

Matteo Bronkhorst did his Bachelor thesis at the pervasive systems group in 2021. His hypothesis was that keeping some sort of attention when using machine learning to learn handwritten characters would be really useful, and he was right.

Applying his findings to the Ubicomp 2021 Challenge organised by Stabilo he has managed to prove that it is possible to achieve great results on IMU-only handwriting recognition by applying a novel Transformer architecture. His results show how much more performing Transformers are with respect to the more traditional RNN and CNN solutions.

Congratulations Matteo on your project and your achievement!