D2-Net: A Trainable CNN for Joint Description and Detection of Local Features
Paper in proceedings, 2019
In this work we address the problem of finding reliable pixel-level correspondences under difficult imaging conditions. We propose an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector. By postponing the detection to a later stage, the obtained keypoints are more stable than their traditional counterparts based on early detection of low-level structures. We show that this model can be trained using pixel correspondences extracted from readily available large-scale SfM reconstructions, without any further annotations. The proposed method obtains state-of-the-art performance on both the difficult Aachen Day-Night localization dataset and the InLoc indoor localization benchmark, as well as competitive performance on other benchmarks for image matching and 3D reconstruction.
deep learning
local features
3D reconstruction
machine learning
visual localization
Author
Mihai Dusmanu
Ecole Normale Superieure (ENS)
INRIA Paris
ETH Zurich
Ignacio Rocco
Ecole Normale Superieure (ENS)
INRIA Paris
Tomas Pajdla
Czech Technical University in Prague
Marc Pollefeys
ETH Zurich
Microsoft Mixed Real & Artificial Intelligence La
Josef Sivic
Ecole Normale Superieure (ENS)
INRIA Paris
Czech Technical University in Prague
Akihiko Torii
Tokyo Institute of Technology
Torsten Sattler
Chalmers, Electrical Engineering, Signal Processing and Biomedical Engineering, Imaging and Image Analysis
2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
8084-8093
IEEE / CVF Conference on Computer Vision and Pattern Recognition
Long Beach, USA,
Long Beach, USA,
Subject Categories
Robotics
Computer Vision and Robotics (Autonomous Systems)
DOI
10.1109/CVPR.2019.00828