GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization

Von Stumberg L, Wenzel P, Khan Q, Cremers D (2020)


Publication Type: Journal article

Publication year: 2020

Journal

Book Volume: 5

Pages Range: 890-897

Article Number: 8954808

Journal Issue: 2

DOI: 10.1109/LRA.2020.2965031

Abstract

Direct SLAM methods have shown exceptional performance on odometry tasks. However, they are susceptible to dynamic lighting and weather changes while also suffering from a bad initialization on large baselines. To overcome this, we propose GN-Net: a network optimized with the novel Gauss-Newton loss for training weather invariant deep features, tailored for direct image alignment. Our network can be trained with pixel correspondences between images taken from different sequences. Experiments on both simulated and real-world datasets demonstrate that our approach is more robust against bad initialization, variations in day-time, and weather changes thereby outperforming state-of-the-art direct and indirect methods. Furthermore, we release an evaluation benchmark for relocalization tracking against different types of weather. Our benchmark is available at https://vision.in.tum.de/gn-net.

Involved external institutions

How to cite

APA:

Von Stumberg, L., Wenzel, P., Khan, Q., & Cremers, D. (2020). GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization. IEEE Robotics and Automation Letters, 5(2), 890-897. https://doi.org/10.1109/LRA.2020.2965031

MLA:

Von Stumberg, Lukas, et al. "GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization." IEEE Robotics and Automation Letters 5.2 (2020): 890-897.

BibTeX: Download