Szczegóły publikacji
Opis bibliograficzny
Well convergent and computationally efficient quaternion loss / Kamil LELOWICZ, Jakub Derbisz // W: Advanced, contemporary control : proceedings of KKA 2020 – the 20th Polish control conference : [14-16 October, 2020], Łódź, Poland / eds. Andrzej Bartoszewicz, Jacek Kabziński, Janusz Kacprzyk. — Cham : Springer Nature Switzerland AG, cop. 2020. — (Advances in Intelligent Systems and Computing ; ISSN 2194-5357 ; vol. 1196). — ISBN: 978-3-030-50935-4; e-ISBN: 978-3-030-50936-1. — S. 1275–1286. — Bibliogr. s. 1285-1286, Abstr. — Publikacja dostępna online od: 2020-06-24. — K. Lelowicz - dod. afiliacja: APTIV Services Poland S. A.
Autorzy (2)
- AGHLelowicz Kamil
- Derbisz Jakub
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 129305 |
|---|---|
| Data dodania do BaDAP | 2020-07-16 |
| DOI | 10.1007/978-3-030-50936-1_106 |
| Rok publikacji | 2020 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Springer |
| Czasopismo/seria | Advances in Intelligent Systems and Computing |
Abstract
Rotation estimation, i.e. the ability to predict angles describing 3D positioned object, is an omnipresent problem in computer vision, computer graphics and 3D object detection task in automotive industry. Deep learning algorithms usually parameterise rotation using only the yaw angle. This paper presents detailed comparison of several 3D rotation distance functions using quaternions. We propose a computationally efficient quaternion loss function for neural network training. We conclude that function respects the topology of SO(3) and is bi-invariant. We also show the geometrical representation of presented functions. Lastly, we evaluate the effectiveness of the proposed loss function and compare its performance with other methods using a public large-scale dataset.
