| Title | A polite robot: visual handshake recognition using deep learning |
|---|---|
| Authors | Butkus, Liutauras ; Lukoševičius, Mantas |
| Full Text |
|
| Is Part of | CEUR workshop proceedings: System 2018: Symposium for young scientists in technology, engineering and mathematics: proceedings of the symposium for young scientists in technology, engineering and mathematics, Gliwice, Poland, May 28, 2018 / edited by G. Capizzi, R. Damaševičius, A. Lopata, T. Krilavičius, Ch. Napoli, M. Woźniak.. Aachen : CEUR-WS. 2018, vol. 2147, p. 78-83.. ISSN 1613-0073 |
| Keywords [eng] | computer vision ; convolutional neural networks ; deep learning ; image recognition ; robotics |
| Abstract [eng] | Our project was to create a demo system where a small humanoid robot accepts an offered handshake when it sees it. The visual handshake recognition, which is the main part of the system proved to be not an easy task. Here we describe how and how well we solved it using deep learning. In contrast to most gesture recognition research we did not use depth information or videos, but did this on static images. We wanted to use a simple camera and our gesture is rather static. We have collected a special dataset for this task. Different configurations and learning algorithms of convolutional neural networks were tried. However, the biggest breakthrough came when we could eliminate the background and make the model concentrate on the person in front. In addition to our experiment results we can also share our dataset. |
| Published | Aachen : CEUR-WS |
| Type | Conference paper |
| Language | English |
| Publication date | 2018 |
| CC license |
|