Unsupervised Deep Homography: A Fast and Robust Homography Estimation Model

By Ty Nguyen, Steven W. Chen, Shreyas S. Shivakumar, Camillo Jose Taylor, Vijay R. Kumar. IEEE Robotics and Automation Letters, 2017

355 citations recorded by the source corpus.

Homography estimation between multiple aerial images can provide relative pose estimation for collaborative autonomous exploration and monitoring. The usage on a robotic system requires a fast and robust homography estimation algorithm. In this letter, we propose an unsupervised learning algorithm that trains a deep convolutional neural network to estimate planar homographies. We compare the proposed algorithm to traditional-feature-based and direct methods, as well as a corresponding supervised learning algorithm. Our empirical results demonstrate that compared to traditional approaches, the unsupervised algorithm achieves faster inference speed, while maintaining comparable or better accuracy and robustness to illumination variation. In addition, our unsupervised method has superior adaptability and performance compared to the corresponding supervised deep learning method. Our image dataset and a Tensorflow implementation of our work are available at https://github.com/tynguyen/unsupervisedDeepHomographyRAL2018.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/5409587af979a088d6d77282c1554a6f68034f11

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

1. Zixin Luo
2. Tian Fang
3. Shiwei Li
4. Yao Yao
5. Long Quan
6. Lei Zhou
7. Hongkai Chen
8. Xuyang Bai

In Computer Vision and Pattern Recognition

1. Tian Fang
2. Zixin Luo
3. Shiwei Li
4. Yao Yao
5. Long Quan
6. Lei Zhou
7. Hongkai Chen
8. Xuyang Bai

Citations in the corpus

• ASLFeat: Learning Local Features of Accurate Shape and Localization (2020, methodology): “Homography-constrained DCN. Virtually, the local deformation can be better approximated by a homography (perspective) transformation H , and we here adopt the Tensor Direct Linear Transform (Tensor DLT) [26] to solve the 4-point parameterization of H in a differentiable manner.”