imclab / sal

Space-Time Saliency for CVPR 2014

Introduction

This page contains software and instructions for Space-Time Saliency [1].

Installation

The code is divided into two parts: Part 1 is hosted on Microsoft's website, containing the core implementation of the space-time saliency method; Part 2 is hosted on GitHub, containing the auxiliary files (including library, toolbox, a video and a demo) to use this code. You need to download both parts and unzip them in the same folder (eg, ./sal).

1. set sal to your current folder in Matlab;
2. Run make.m in Matlab to compile all C++ files;
3. Run addPath.m to add sub-directories into the path of Matlab.
4. Run demoSal.m file.

Instructions

The package contains the following files and folders:

• ./data: This folder contains a video sequence as example.
• ./core: This folder contains the main implementation of the space-time saliency algorithm.
• ./src: This folder contains a wrapper of the space-time saliency for any video input.
• ./lib: This folder contains some necessary library functions.
• ./tool: This folder contains some 3rd party toolboxes.
• ./make.m: Matlab makefile for C++ code.
• ./addPath.m: Adds the sub-directories into the path of Matlab.
• ./demoSal.m: A demo file for generating and visualizing the saliency for a video input.

FAQs

References

[1] F. Zhou, S.-B. Kang, and M. Cohen, "Time-Mapping Using Space-Time Saliency," in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014

Copyright

This software is free for use in research projects. If you publish results obtained using this software, please use this citation.

@inproceedings{ZhouKC14,
author    = {F. Zhou and S.-B. Kang and M. Cohen},
title     = {Time-Mapping Using Space-Time Saliency},
booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year      = {2014},
}

If you have any question, please feel free to contact Feng Zhou ([email protected]).