Movement-induced priors for deep stereo

Yuxin Hou, Muhammad Kamran Janjua, Juho Kannala, Arno Solin

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussaConference article in proceedingsScientificvertaisarvioitu

Abstrakti

We propose a method for fusing stereo disparity estimation with movement-induced prior information. Instead of independent inference frame-by-frame, we formulate the problem as a non-parametric learning task in terms of a temporal Gaussian process prior with a movement-driven kernel for inter-frame reasoning. We present a hierarchy of three Gaussian process kernels depending on the availability of motion information, where our main focus is on a new gyroscope-driven kernel for handheld devices with low-quality MEMS sensors, thus also relaxing the requirement of having full 6D camera poses available. We show how our method can be combined with two state-of-the-art deep stereo methods. The method either work in a plug-and-play fashion with pre-trained deep stereo networks, or further improved by jointly training the kernels together with encoder-decoder architectures, leading to consistent improvement.

AlkuperäiskieliEnglanti
OtsikkoProceedings of ICPR 2020 - 25th International Conference on Pattern Recognition
KustantajaIEEE
Sivut7478-7485
Sivumäärä8
ISBN (elektroninen)9781728188089
DOI - pysyväislinkit
TilaJulkaistu - 2020
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisussa
TapahtumaInternational Conference on Pattern Recognition - Virtual, Online, Milan, Italia
Kesto: 10 tammik. 202115 tammik. 2021
Konferenssinumero: 25

Julkaisusarja

NimiProceedings - International Conference on Pattern Recognition
ISSN (painettu)1051-4651

Conference

ConferenceInternational Conference on Pattern Recognition
LyhennettäICPR
Maa/AlueItalia
KaupunkiMilan
Ajanjakso10/01/202115/01/2021

Sormenjälki

Sukella tutkimusaiheisiin 'Movement-induced priors for deep stereo'. Ne muodostavat yhdessä ainutlaatuisen sormenjäljen.

Siteeraa tätä