Towards CNN map representation and compression for camera relocalisation

Luis A Contreras, Walterio Mayol-Cuevas

Research output: Contribution to conferenceConference Paperpeer-review

113 Downloads (Pure)

Abstract

This paper presents a study on the use of Convolutional Neural Networks for camera relocalisation and its application to map compression.We follow state of the art visual relocalisation results and evaluate the response to different data inputs. We use a CNN map representation and introduce the notion of map compression under this paradigm by using smaller CNN architectures without sacrificing relocalisation performance. We evaluate this approach in a series of publicly available datasets over a number of CNN architectures with different sizes, both in complexity and number of layers. This formulation allows us to improve relocalisation accuracy by increasing the number of training trajectories while maintaining a constant-size CNN.
Original languageEnglish
Number of pages8
Publication statusPublished - 18 Jun 2018
Event1st International Workshop on Deep Learning for VisuAL SLAM - Salt Lake City, United States
Duration: 18 Jun 2018 → …

Conference

Conference1st International Workshop on Deep Learning for VisuAL SLAM
CountryUnited States
CitySalt Lake City
Period18/06/18 → …

Fingerprint Dive into the research topics of 'Towards CNN map representation and compression for camera relocalisation'. Together they form a unique fingerprint.

Cite this