G2MF-WA: Geometric multi-model fitting with weakly annotated data
- PDF / 2,800,174 Bytes
- 11 Pages / 612 x 808 pts Page_size
- 90 Downloads / 199 Views
Vol. 6, No. 2, June 2020, 135–145
Research Article
G2MF-WA: Geometric multi-model fitting with weakly annotated data Chao Zhang1 ( ), Xuequan Lu2 , Katsuya Hotta1 , and Xi Yang3 c The Author(s) 2020.
Abstract In this paper we address the problem of geometric multi-model fitting using a few weakly annotated data points, which has been little studied so far. In weak annotating (WA), most manual annotations are supposed to be correct yet inevitably mixed with incorrect ones. Such WA data can naturally arise through interaction in various tasks. For example, in the case of homography estimation, one can easily annotate points on the same plane or object with a single label by observing the image. Motivated by this, we propose a novel method to make full use of WA data to boost multi-model fitting performance. Specifically, a graph for model proposal sampling is first constructed using the WA data, given the prior that WA data annotated with the same weak label has a high probability of belonging to the same model. By incorporating this prior knowledge into the calculation of edge probabilities, vertices (i.e., data points) lying on or near the latent model are likely to be associated and further form a subset or cluster for effective proposal generation. Having generated proposals, α-expansion is used for labeling, and our method in return updates the proposals. This procedure works in an iterative way. Extensive experiments validate our method and show that it produces noticeably better results than state-of-the-art techniques in most cases.
1
Introduction
Geometric model fitting aims to fit a model to data which contains both inliers and outliers. A wellknown approach is RANSAC [1], the main idea of which is to generate a number of random model proposals and select as the best solution the one which includes the largest inlier set based on an inlier threshold. The geometric multi-model fitting task further assumes that the input data requires multiple models. Multi-model fitting algorithms have to optimize the global solution, rather than taking a greedy strategy to maximize inliers for single models like RANSAC. To evaluate the numerous possible solutions, a common approach is to design an energy function [2–5], such that an approximate solution can be achieved by energy minimization (or maximization) by balancing geometric errors (data fidelity) and regularity of inlier clusters (e.g., smoothness, complexity). Although finding the optimal solution is NP-hard [4], α-expansion [2] provides a powerful alternative which can find solutions with guaranteed approximation bounds over a given set of model proposals. However, the quality of the solution and convergence largely depend on the quality of the proposals, which greatly influence Keywords geometric multi-model fitting; weak annota- the overall efficiency and effectiveness. tion; multi-homography detection; two-view Most methods attempt to improve the quality of motion segmentation model proposals by sampling “clean” subsets of data points from the input data. We, however,
Data Loading...
