Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Latent Semantic Consensus For Deterministic Geometric Model Fitting

Guobao Xiao, Jun Yu, Jiayi Ma, Deng-Ping Fan, Ling Shao

TL;DR

This paper attempts to sample high-quality subsets and select model instances to estimate parameters in the multi-structural data and proposes an effective method called Latent Semantic Consensus (LSC), able to provide consistent and reliable solutions for general multi-structural model fitting.

Abstract

Estimating reliable geometric model parameters from the data with severe outliers is a fundamental and important task in computer vision. This paper attempts to sample high-quality subsets and select model instances to estimate parameters in the multi-structural data. To address this, we propose an effective method called Latent Semantic Consensus (LSC). The principle of LSC is to preserve the latent semantic consensus in both data points and model hypotheses. Specifically, LSC formulates the model fitting problem into two latent semantic spaces based on data points and model hypotheses, respectively. Then, LSC explores the distributions of points in the two latent semantic spaces, to remove outliers, generate high-quality model hypotheses, and effectively estimate model instances. Finally, LSC is able to provide consistent and reliable solutions within only a few milliseconds for general multi-structural model fitting, due to its deterministic fitting nature and efficiency. Compared with several state-of-the-art model fitting methods, our LSC achieves significant superiority for the performance of both accuracy and speed on synthetic data and real images. The code will be available at https://github.com/guobaoxiao/LSC.

Latent Semantic Consensus For Deterministic Geometric Model Fitting

TL;DR

This paper attempts to sample high-quality subsets and select model instances to estimate parameters in the multi-structural data and proposes an effective method called Latent Semantic Consensus (LSC), able to provide consistent and reliable solutions for general multi-structural model fitting.

Abstract

Estimating reliable geometric model parameters from the data with severe outliers is a fundamental and important task in computer vision. This paper attempts to sample high-quality subsets and select model instances to estimate parameters in the multi-structural data. To address this, we propose an effective method called Latent Semantic Consensus (LSC). The principle of LSC is to preserve the latent semantic consensus in both data points and model hypotheses. Specifically, LSC formulates the model fitting problem into two latent semantic spaces based on data points and model hypotheses, respectively. Then, LSC explores the distributions of points in the two latent semantic spaces, to remove outliers, generate high-quality model hypotheses, and effectively estimate model instances. Finally, LSC is able to provide consistent and reliable solutions within only a few milliseconds for general multi-structural model fitting, due to its deterministic fitting nature and efficiency. Compared with several state-of-the-art model fitting methods, our LSC achieves significant superiority for the performance of both accuracy and speed on synthetic data and real images. The code will be available at https://github.com/guobaoxiao/LSC.
Paper Structure (28 sections, 28 equations, 19 figures, 6 tables, 2 algorithms)

This paper contains 28 sections, 28 equations, 19 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Random Model Fitting Methods
  4. Deterministic Model Fitting Methods
  5. Proposed Method
  6. Problem Formulation
  7. Latent Semantic Analysis for Model Fitting
  8. Latent Semantic Consensus based Sampling Algorithm
  9. Latent Semantic Consensus based Model Selection Algorithm
  10. The Complete Method and Analysis
  11. Experiments
  12. Dataset Setting and Evaluation Metrics
  13. Parameter Analysis and Settings
  14. Results on Synthetic Data
  15. Line Fitting
  16. ...and 13 more sections

Figures (19)

  • Figure 1: An example of latent semantic space for line fitting. (a) The input data with data points and the model hypotheses generated by the proposed sampling method. (b) The distribution of inliers and model hypotheses corresponding to different model instances (i.e., S1, S2 and S3). Others are gross outliers and bad model hypotheses. (c) and (d) The latent semantic space of data points and model hypotheses. The gross outliers and bad model hypotheses are marked in blue, and the inliers and good model hypotheses are marked in different color. The red point is the origin of latent semantic space.
  • Figure 2: An example of the distance distribution from inliers (red) and outliers (blue) in DP-LSS.
  • Figure 3: An example of computing the residual value between an origin-line $\vec{l}_{\hat{\theta}_j}$ and a point $\hat{\theta}_i$ in MH-LSS. The residual value is the Euclidean distance pointed by a red arrow.
  • Figure 4: Quantitative results obtained by the proposed LSC fitting scheme with different values of $\psi$ for the task of (Top) homography estimation and (Bottom) fundamental matrix estimation on the MS-COCO-F, YFCC100M-F and AdelaideRMF datasets.
  • Figure 5: Quantitative results obtained by the proposed LSC fitting scheme with different values of $\beta$ for the task of homography estimation and fundamental matrix estimation on the AdelaideRMF dataset.
  • ...and 14 more figures