Unsupervised learning for robust fitting: A reinforcement learning approach

Abstract

Robust model fitting is a core algorithm in a large number of computer vision applications. Solving this problem efficiently for datasets highly contaminated with outliers is, however, still challenging due to the underlying computational complexity. Recent literature has focused on learning-based algorithms. However, most approaches are supervised (which require a large amount of labelled training data). In this paper, we introduce a novel unsupervised learning framework that learns to directly solve robust model fitting. Unlike other methods, our work is agnostic to the underlying input features, and can be easily generalized to a wide variety of LP-type problems with quasi-convex residuals. We empirically show that our method out-performs existing unsupervised learning approaches, and achieves competitive results compared to traditional methods on several important computer vision problems

RAS ID

36862

Document Type

Conference Proceeding

Date of Publication

2021

Funding Information

Australian Research Council

Wallenberg AI

Autonomous Systems and Software Program Chalmers

AI Research Center (CHAIR) Seed Projects

School

School of Science

Grant Number

ARC Number : DP200103448

Copyright

free_to_read

Publisher

IEEE

Comments

Truong, G., Le, H., Suter, D., Zhang, E., & Gilani, S. Z. (2021). Unsupervised learning for robust fitting: A reinforcement learning approach. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 10343-10352). IEEE.

https://doi.org/10.1109/CVPR46437.2021.01021

An open access version of this paper is provided by the Computer Vision Foundation.

Share

 
COinS
 

Link to publisher version (DOI)

10.1109/CVPR46437.2021.01021