UnsuperPoint: End-to-end Unsupervised Interest Point Detector and Descriptor

Publikation: Working paperForskning

Standard

UnsuperPoint : End-to-end Unsupervised Interest Point Detector and Descriptor. / Christiansen, Peter Hviid; Kragh, Mikkel Fly; Brodskiy, Yury; Karstoft, Henrik.

arXiv.org, 2019.

Publikation: Working paperForskning

Harvard

APA

CBE

MLA

Christiansen, Peter Hviid o.a.. UnsuperPoint: End-to-end Unsupervised Interest Point Detector and Descriptor. arXiv.org. 2019., 14 s.

Vancouver

Author

Bibtex

@techreport{bed31da3cb564dfea0823bfa8d8f08b3,
title = "UnsuperPoint: End-to-end Unsupervised Interest Point Detector and Descriptor",
abstract = "It is hard to create consistent ground truth data for interest points in natural images, since interest points are hard to define clearly and consistently for a human annotator. This makes interest point detectors non-trivial to build. In this work, we introduce an unsupervised deep learning-based interest point detector and descriptor. Using a self-supervised approach, we utilize a siamese network and a novel loss function that enables interest point scores and positions to be learned automatically. The resulting interest point detector and descriptor is UnsuperPoint. We use regression of point positions to 1) make UnsuperPoint end-to-end trainable and 2) to incorporate non-maximum suppression in the model. Unlike most trainable detectors, it requires no generation of pseudo ground truth points, no structure-from-motion-generated representations and the model is learned from only one round of training. Furthermore, we introduce a novel loss function to regularize network predictions to be uniformly distributed. UnsuperPoint runs in real-time with 323 frames per second (fps) at a resolution of 224x320 and 90 fps at 480x640. It is comparable or better than state-of-the-art performance when measured for speed, repeatability, localization, matching score and homography estimation on the HPatch dataset.",
keywords = "Deep Learning, Interest Point Detector, Point Descriptor, Point detector, Unsupervised, Self-supervised, Real-time",
author = "Christiansen, {Peter Hviid} and Kragh, {Mikkel Fly} and Yury Brodskiy and Henrik Karstoft",
year = "2019",
language = "English",
publisher = "arXiv.org",
type = "WorkingPaper",
institution = "arXiv.org",

}

RIS

TY - UNPB

T1 - UnsuperPoint

T2 - End-to-end Unsupervised Interest Point Detector and Descriptor

AU - Christiansen, Peter Hviid

AU - Kragh, Mikkel Fly

AU - Brodskiy, Yury

AU - Karstoft, Henrik

PY - 2019

Y1 - 2019

N2 - It is hard to create consistent ground truth data for interest points in natural images, since interest points are hard to define clearly and consistently for a human annotator. This makes interest point detectors non-trivial to build. In this work, we introduce an unsupervised deep learning-based interest point detector and descriptor. Using a self-supervised approach, we utilize a siamese network and a novel loss function that enables interest point scores and positions to be learned automatically. The resulting interest point detector and descriptor is UnsuperPoint. We use regression of point positions to 1) make UnsuperPoint end-to-end trainable and 2) to incorporate non-maximum suppression in the model. Unlike most trainable detectors, it requires no generation of pseudo ground truth points, no structure-from-motion-generated representations and the model is learned from only one round of training. Furthermore, we introduce a novel loss function to regularize network predictions to be uniformly distributed. UnsuperPoint runs in real-time with 323 frames per second (fps) at a resolution of 224x320 and 90 fps at 480x640. It is comparable or better than state-of-the-art performance when measured for speed, repeatability, localization, matching score and homography estimation on the HPatch dataset.

AB - It is hard to create consistent ground truth data for interest points in natural images, since interest points are hard to define clearly and consistently for a human annotator. This makes interest point detectors non-trivial to build. In this work, we introduce an unsupervised deep learning-based interest point detector and descriptor. Using a self-supervised approach, we utilize a siamese network and a novel loss function that enables interest point scores and positions to be learned automatically. The resulting interest point detector and descriptor is UnsuperPoint. We use regression of point positions to 1) make UnsuperPoint end-to-end trainable and 2) to incorporate non-maximum suppression in the model. Unlike most trainable detectors, it requires no generation of pseudo ground truth points, no structure-from-motion-generated representations and the model is learned from only one round of training. Furthermore, we introduce a novel loss function to regularize network predictions to be uniformly distributed. UnsuperPoint runs in real-time with 323 frames per second (fps) at a resolution of 224x320 and 90 fps at 480x640. It is comparable or better than state-of-the-art performance when measured for speed, repeatability, localization, matching score and homography estimation on the HPatch dataset.

KW - Deep Learning

KW - Interest Point Detector

KW - Point Descriptor

KW - Point detector

KW - Unsupervised

KW - Self-supervised

KW - Real-time

M3 - Working paper

BT - UnsuperPoint

PB - arXiv.org

ER -