Vanishing Points detection

version 0.9 - June 2017 by Jose Lezama jlezama@fing.edu.uy

Introduction

This code implements the vanishing point detection algorithm as described in the CVPR 2014 article "Finding vanishing points via point alignments in image primal and dual domains", Jose Lezama, Rafael Grompone von Gioi, Gregory Randall and Jean-Michel Morel, and IPOL article "Vanishing Point Detection in Urban Scenes Using Point Alignments" Jose Lezama, Gregory Randall, Jean-Michel Morel and Rafael Grompone von Gioi.

Optionally, this code uses the algorithm by Figueiredo and Jain, Unsupervised Learning of Finite Mixture Models, TPAMI 2002, to quickly obtain cluster candidates.

Files and Folders

• README.txt - This file
• COPYING - GNU AFFERO GENERAL PUBLIC LICENSE Version 3
• Makefile - Compilation instructions for 'make'
• main.m - demo script
• detect_vps.m - main algorithm script
• yud_benchmark.m - script to run benchmark scores in York Urban Database
• ecd_benchmark.m - script to run benchmark scores in Eurasian Citites Database
• test.jpg - test image
• lib/ - folder with auxiliary MATLAB scripts
• mex_files/ - folder with C sources for line segment and point alignment detection, to be compiled as mex files
• mixtures/ - folder with the unsupervised mixtures detection code of Figueiredo and Jain, which can optionally be used for accelerating the method

Compiling

The algorithm depends on three mex scripts that need to be compiled before execution. For compilation inside MATLAB, cd into the mex_files folder and run build.m

Optional: run make to produce compiled Matlab executables

Running

For a test run on the test image, run main.m main.m calls the main function, detect_vps.m Arguments of detect_vps.m are:

• img_in: filename of the input image
• folder_out: path to save resulting image and text files
• manhattan: boolean variable used to determine if the Manhattan-world hypothesis is assumed
• acceleration: boolean variable used to determine if acceleration using Figueiredo and Jain GMM algorithm should be used
• focal_ratio: ratio between the focal lenght and captor width
• input_params: optional input parameters

Benchmarks

To run benchmarks on York Urban Dataset (YUD) and Eurasian Cities Dataset (ECD) run yud_benchmark.m and ecd_benchmark.m. You should obtain results similar or better to the ones reported in our CVPR paper.

Note that without the acceleration the scripts can be slow. In particular for ECD the non-accelerated version can take up to 3 minutes per image (there are 103 images in that dataset).

The datasets are not provided but can be obtained from the following sites:

YUD, by P. Denis: http://www.elderlab.yorku.ca/YorkUrbanDB/ ECD, by O. Barinova: http://graphics.cs.msu.ru/en/research/projects/msr/geometry

Acceleration

As an optional procedure, an accelerated version of the algorithm can be run by setting the appropiate flag (see detect_vps.m). This version uses Figueiredo's "Unsupervised Learning of Finite Mixture Models" algorithm to quickly obtain cluster candidates.

The scripts files, available at the author's website (http://www.lx.it.pt/~mtf/) are included, and have been slightly modified for speed improvement.

Copyright and License

Copyright (c) 2013-2015 Jose Lezama jlezama@fing.edu.uy

This program is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.

You should have received a copy of the GNU Affero General Public License along with this program. If not, see http://www.gnu.org/licenses/.

Thanks

We would be grateful to receive any comment, especially about errors, bugs, or strange results.