Content Tags

There are no tags.

Unsupervised Deep Homography: A Fast and Robust Homography Estimation Model.

RSS Source
Authors
Ty Nguyen, Steven W. Chen, Shreyas S. Shivakumar, Camillo J. Taylor, Vijay Kumar

Homography estimation between multiple aerial images can provide relativepose estimation for collaborative autonomous exploration and monitoring. Theusage on a robotic system requires a fast and robust homography estimationalgorithm. In this study, we propose an unsupervised learning algorithm thattrains a Deep Convolutional Neural Network to estimate planar homographies. Wecompare the proposed algorithm to traditional feature-based and direct methods,as well as a corresponding supervised learning algorithm. Our empirical resultsdemonstrate that compared to traditional approaches, the unsupervised algorithmachieves faster inference speed, while maintaining comparable or betteraccuracy and robustness to illumination variation. In addition, on both asynthetic dataset and representative real-world aerial dataset, ourunsupervised method has superior adaptability and performance compared to thesupervised deep learning method.

Stay in the loop.

Subscribe to our newsletter for a weekly update on the latest podcast, news, events, and jobs postings.