Scene Coordinate Regression with Angle-Based Reprojection Loss for Camera Relocalization

Xiaotian Li, Juha Ylioinas, Jakob Verbeek, Juho Kannala

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review


Image-based camera relocalization is an important problem in computer vision and robotics. Recent works utilize convolutional neural networks (CNNs) to regress for pixels in a query image their corresponding 3D world coordinates in the scene. The final pose is then solved via a RANSAC-based optimization scheme using the predicted coordinates. Usually, the CNN is trained with ground truth scene coordinates, but it has also been shown that the network can discover 3D scene geometry automatically by minimizing single-view reprojection loss. However, due to the deficiencies of the reprojection loss, the network needs to be carefully initialized. In this paper, we present a new angle-based reprojection loss, which resolves the issues of the original reprojection loss. With this new loss function, the network can be trained without careful initialization, and the system achieves more accurate results. The new loss also enables us to utilize available multi-view constraints, which further improve performance.
Original languageEnglish
Title of host publicationComputer Vision – ECCV 2018 Workshops
Subtitle of host publicationMunich, Germany, September 8-14, 2018, Proceedings, Part III
ISBN (Electronic)978-3-030-11015-4
ISBN (Print)978-3-030-11014-7
Publication statusPublished - 2019
MoE publication typeA4 Conference publication
EventEuropean Conference on Computer Vision - Munich, Germany
Duration: 8 Sept 201814 Sept 2018
Conference number: 15

Publication series

NameLecture Notes in Computer Science


ConferenceEuropean Conference on Computer Vision
Abbreviated titleECCV


Dive into the research topics of 'Scene Coordinate Regression with Angle-Based Reprojection Loss for Camera Relocalization'. Together they form a unique fingerprint.

Cite this