PG-framework

Introduction

This is a perceptual grouping framework that organizes image edges into meaningful structures.

If you use the toolbox, we appreciate it if you cite our paper:

@inproceedings{DBLP:conf/cvpr/QiSXZHLG15, author = {Yonggang Qi and Yi{-}Zhe Song and Tao Xiang and Honggang Zhang and Timothy M. Hospedales and Yi Li and Jun Guo}, title = {Making better use of edges via perceptual grouping}, booktitle = {{IEEE} Conference on Computer Vision and Pattern Recognition, {CVPR} 2015, Boston, MA, USA, June 7-12, 2015}, pages = {1856--1865}, year = {2015}}

System Requirements

• Tested on Matlab R2013b for 64-bit, Ubuntu OS.
• Note that there is a link error to gco toolbox on Matlab R2014 or above, so please make sure run this code on Matlab R2013.

---

Installation

• This package should contain the following files:

  • Main code

    • PGDemo.m : The Demo usage of perceptual edge grouping.
    • matlab*.m : The Matlab files of implementation of the corresponding paper mentioned in Introduction.

  • Third party toolbox

    • EdgeDetector\ : StructuredEdge detector for edge detection.
    • Edge Linking and Line Segment Fitting\ : Edge cutting toolbox for obtaining segments on edge map.
    • piotr_toolbox_V3.02\ : Piotr's Matlab toolbox.
    • gco-v3.0\ : software for energy minimization with graph cuts

  • Data

    • images/*.jpg : Exemple input pictures.

• Start Matlab and add all the above files to Path.

---

Getting started

• Make sure to carefully follow the installation instructions above.
• Please see "PGDemo.m" to run demo and get basic usage information.

---

History

Version 1.0 (30/05/2016)

• initial version