D2-Net: A Trainable CNN for Joint Detection and Description of Local Features

Authors:
  • MDMihai Dusmanu
    Mihai Dusmanu
  • IRIgnacio Rocco
    Ignacio Rocco
  • TPTomas Pajdla
    Tomas Pajdla
  • MPMarc Pollefeys
    Marc Pollefeys
  • JSJosef Sivic
    Josef Sivic
  • ATAkihiko Torii
    Akihiko Torii
  • TSTorsten Sattler
    Torsten Sattler
Preprint
Publisher website
Google Scholar
Abstract
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.
All Related Versions