telecommunication-telemedia-assessment / gbvs360-bms360-prosal

Extending existing saliency prediction models from 2D to omnidirectional images

GBVS360, BMS360, ProSal: Extending existing saliency prediction models from 2D to omnidirectional images

In this project, it is studied how existing visual saliency models designed for 2D images can be extended and applied to Omnidirectional images in the equirectangular format. Three different frameworks, BMS360, GBVS360 and Projected Saliency (ProSal) were designed to address this task. In the particular case of BMS360 and GBVS360, the 2D models Boolean Map Saliency (BMS) and Graph-Based Visual Saliency (GBVS) were rigorously extended to integrate the specific properties of equirectangular images to model visual attention in omnidirectional images watched using Head Mounted Displays (HMDs). With the proposed extensions, the saliency prediction performance in comparison to their original design could significantly be improved.

Context

This project was intiated to address the ICME2017 Grand Challenge on Visual attention for 360 images. Two track of this challenged are considered:

• The prediction of saliency maps obtained from Head motion data (on still images)

• The prediction of saliency maps obtained from Head and Eye motion data (on still images)

Binaries

In the release section of GitHub, different set of binaries can be found. The version 1.0 correspond to the models submitted to the ICME2017 Grand Challenge. These models were ranked as 1st at addressing Head motion-based saliency maps out of 13 submissions. And was ranked and was ranked 6th (5.75) out of 16 submissions when Head/Eye motion-based saliency maps are addressed.

The release with a version higher than 1.0 were the result of work performed after the end of the competition. Version 2.0 includes the BMS360 model.

Build from source

To compile the project, only two dependencies are required:

• fftw (optional: it adds support of a CSF extention to GBVS)

• boost (>= 1.63)

• OpenCV (>= 3.0)

UNIX system

run make.

Windows:

A visual studio solution file is provided (tested using the 2015 Community edition). A complete set of all depending libraries can be found at the following URL: https://1drv.ms/u/s!AqL868MBKJ4Gh-guxlbtLEQ4h6VqXw Unzip the folder Library next to the clone of the repository. Then, open the visual studio solution file, and you can compile the project (The path to the external library folder is set relatively to the root folder).

Citation

The journal paper related to this work has been accepted for publication. In case you use this work, please cite:

@article{lebreton2018GBVS360,
  title={GBVS360, BMS360, ProSal: Extending existing saliency prediction models from 2D to omnidirectional images},
  author={Lebreton, Pierre and Raake, Alexander},
  journal={Signal Processing: Image Communication},
  volume={69},
  pages={69--78},
  year={2018},
  month={november},
  publisher={Elsevier}
}

For the evaluation scripts used in this study, please refer and cite:

@article{Gutierrez2018Toolbox, 
  title={Toolbox and Dataset for the Development of Saliency and Scanpath Models for Omnidirectional / 360{$^{\circ}$} Still Images}, 
  author={Guti{\'e}rrez, Jes{\'u}s and David, Erwan and Rai, Yashas and Le Callet, Patrick 
}, 
  journal={Signal Processing: Image Communication}, 
  volume={69}, 
  pages={35--42}, 
  month={november},
  year={2018}, 
  publisher={Elsevier} 
}

License

The MIT License (MIT)

Copyright (c) 2017 Pierre Lebreton

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

gbvs360-bms360-prosal's Issues

Processing 360 video

Hello,
I have read on the README from tag 2.0 that the model is able to extract the saliency from videos this way:

The tool can also support videos with a large variety of codec and containers. The output saliency map will be another video encoded in H.264 in a .mkv containter. To process a video, use the option -i and provide the video to process as parameter.

I have tried to do this to an mp4 360 video called VR by typing .\salient.exe -i VR.mp4 --general-model 1 -o outputvid.mkv , but got the error

Mayve this is a quick fix, do you know how to proceed?

problem compiling code

Hey,

Thank you for making the GBVS360 code publicly available. I have a question about compiling that I'm wondering if you could help me with. I'm wondering if my path structure is correct.

I unzipped the library folder from the link provided and placed it at the same level of my directory hierarchy as the cloned GBVS360 folder. Specifically, my directory structure looks like this:

C:\Users\Robertson Lab\Desktop\Salience\GBVS360-BMS360-ProSal-master
C:\Users\Robertson Lab\Desktop\Salience\Library

When I load the solution file "Salient360.sln" in Visual Studio Community 2015 and attempt to compile, I receive multiple of the following errors (pasted below). Do you have any advice for compiling the code from here?

Thank you for your help in advance!

GBVS360 with equatorial prior

Hi,

Thank you very much for the nice work and making your models available.
I'm trying to use GBVS360 model with equatorial prior by using the command

salient.exe -i P10.jpg --general-model 2 -o saliency.jpg --legacy-icme-2017

as suggested here : https://github.com/Telecommunication-Telemedia-Assessment/GBVS360-BMS360-ProSal/releases/tag/2.0
However, it looks like the equatorial prior is not effective. When I add the argument for it and run the model with this command it seems fine :
salient.exe -i P10.jpg --general-model 2 -o saliency.jpg --legacy-icme-2017 --equatorial-prior

Here you can find an example image, the output of the first command, and the output of the second command :

Am I misusing the commands ? Or is there a typo in the given command ? I'd be very happy if you can point out what I am missing. Thanks in advance!

Best,
Bahar

opencv_world320.dll

0x00007FFDFB82722A (opencv_world320.dll) (Salient360.exe ) An unhandled exception 0xC0000005

a problem in make libgbvs

Makefile of libgbvs：

CC = clang++
CL = ar
CLFLAGS = rvs

CFLAGS = -W -Wall -g -std=c++14 \
                         -I/usr/local/include \
                         -I../liblinper/src \
                         -I../libmeanshift/src

LDFLAGS = -L/usr/local/lib \
          -L../liblinper/bin \
          -I/usr/local/include \
                  -L../libmeanshift/bin \
                  -lopencv_core -lopencv_imgproc -lopencv_objdetect -lopencv_highgui -lopencv_imgcodecs \
                  -lboost_program_options -lboost_exception -lboost_thread -lboost_system \
                  -lfftw3 \
                  -llinper \
                  -lmean_shift

---

error msg:

clang++ -W -Wall -g -std=c++14 -I/usr/local/include -I../liblinper/src -I../libmeanshift/src  -o src/GBVS.o -c src/GBVS.cpp 
clang++ -W -Wall -g -std=c++14 -I/usr/local/include -I../liblinper/src -I../libmeanshift/src  -o src/main.o -c src/main.cpp 
clang++ -L/usr/local/lib -L../liblinper/bin -I/usr/local/include -I../liblinper/src -L../libmeanshift/bin -lopencv_core -lopencv_imgproc -lopencv_objdetect -lopencv_highgui -lopencv_imgcodecs -lboost_program_options -lboost_exception -lboost_thread -lboost_system -lfftw3 -llinper -lmean_shift -o bin/gbvs src/GBVS.o src/main.o src/fftw++.cc
src/GBVS.o： ‘GBVS::getPerspectiveFeatureJob(cv::Mat const&, cv::Mat&, double*) const’ ：
/home/GBVS360-BMS360-ProSal-master/lib/libgbvs/src/GBVS.cpp:1644： ‘vanishingLineFeatureMapF64C3(cv::Mat const&, double*)’ no define the include
src/GBVS.o：in ‘GBVS::segmentationFeature(cv::Mat const&) const’：
/home/GBVS360-BMS360-ProSal-master/lib/libgbvs/src/GBVS.cpp:1826：‘msseg_f(cv::Mat&, float, float, unsigned int, cv::Mat&)’no define the include
clang: error: linker command failed with exit code 1 (use -v to see invocation)

gnomonic command line interface

Hello,

Thank you again for your help with the GBVS360 libraries! I am currently trying to access the functionality of the gnomonic dir. I am using VirtualBox Ubuntu 16.04 LTS. Here is a picture of my path as well as the "make" inside the dir.

I was wondering if you could provide guidance to building this interface (I assume it can be used to get rectilinear projects from equirectangular maps).

Thank you in advance!

Output image not generating

Thanks for making this code accessible!

I am working on a windows computer and am having trouble getting images to export. As a very simple test, I've moved a test equirectangular image into the \GBVS360-BMS360-ProSal\x64\Debug directory

Using PowerShell in this same directory I am running the command:
.\Salient360.exe -i .\test.jpg -o .\testout.jpg --general-model 2

No errors are generated but no output image in this directory is exported. Am I using this command incorrectly or might there be something else going wrong? Thank you!

How to operate after the project environment is equipped？

I run the code on vs2015. The opencv and boost is installed，but i got "The program '[9960] Salient360.exe' has exited with code 0 (0x0)."
Is there a good person to tell me the next steps in detail? The README is not very detailed