benlaug / labgen-p

Source code of the LaBGen-P stationary background generation method.

LaBGen-P

LaBGen-P is a pixel-level stationary background generation method introduced in [1], and based on LaBGen. Our method has been ranked second during the IEEE Scene Background Modeling Contest (SBMC) 2016. The current ranking and the quantitative metrics computed on the SBMnet dataset are available here.

The purpose of this repository is twofold:

1. To share the source code of the method.
2. To embed the method in a ready-to-use program.

Here is a video showing some backgrounds estimated by LaBGen-P (click on the image below to play it):

Compiling the program

The program implementing the method has been developed in C++11 and is distributed under the GPLv3 license. In order to compile it, you need a modern C++ compiler, a copy of the Boost library, a copy of the OpenCV library, and the CMake build automation tool. On UNIX-like environments, the program can be compiled as follows, considering that your terminal is in the source code directory:

$ cd build
$ cmake -DCMAKE_BUILD_TYPE=Release ..
$ make

Running the program

Once the program has been compiled, the following command gives the complete list of available options:

$ ./LaBGen-P-cli --help

As an example, the IBMtest2 sequence of the SBI dataset [4] can be processed with the default set of parameters as follows:

$ ./LaBGen-P-cli -i path_to_IBMtest2/IBMtest2_%6d.png -o my_output_path -d -v

A full documentation of the options of the program is available on the wiki.

Citation

If you use LaBGen-P in your work, please cite paper [1] as below:

@inproceedings{Laugraud2016LaBGen-P,
  title = {{LaBGen-P}: A Pixel-Level Stationary Background Generation Method Based on {LaBGen}},
  author = {B. Laugraud and S. Pi{\'e}rard and M. {Van Droogenbroeck}},
  booktitle = {IEEE International Conference on Pattern Recognition (ICPR), IEEE Scene Background Modeling Contest (SBMC)},
  pages = {107-113},
  year = {2016},
  month = {December},
  address = {Canc{\'u}n, Mexico},
  doi = {10.1109/ICPR.2016.7899617}
}

Alternatives

• The original patch-based version of LaBGen.
• A variant of LaBGen, called LaBGen-OF, and leverages optical flow algorithms for motion detection.

Testing

Each commited revision is automatically tested using Travis CI on:

• Ubuntu 14.04 with the g++ compiler and OpenCV 2.3 installed from the Ubuntu repositories.
• Ubuntu 14.04 with the g++ compiler and OpenCV 3.2 compiled from the sources.
• OS X El Capitan with the clang++ compiler and OpenCV 2.4 installed with Homebrew.
• OS X El Capitan with the clang++ compiler and OpenCV 3.2 installed with Homebrew.

References

[1] B. Laugraud, S. Piérard, M. Van Droogenbroeck. LaBGen-P: A Pixel-Level Stationary Background Generation Method Based on LaBGen. International Conference on Pattern Recognition (ICPR), IEEE Scene Background Modeling Contest (SBMC), 107–113, 2016.

[2] L. Maddalena, A. Petrosino. Towards Benchmarking Scene Background Initialization. International Conference on Image Analysis and Processing Workshops (ICIAP Workshops), 9281:469-476, 2015.

labgen-p's Issues

Processing sequences ...

It looks like this program reads the entire sequence into memory before doing any processing.
Is that right? If so, its very hard to process video.

Cannot read image sequences

I have the following error message when trying to use LabGen-P with --list_images=../input/in%6d.jpg . The input folder contains image sequences from 'in001160.jpg' -> 'in001384.jpg'

[image2 @ 0x251d1e0] Could find no file with path '../input/in%6d.jpg' and index in the range 0-4

Do you have any idea about this error message? It seems that it is a limitation of OpenCV's VideoCapture object. Thanks in advance!

1. OPENCV_VERSION don't exist in opencv2.4.11 2. #180 in History.cpp missing ns_labgen_p namespace befor Utils

This fist can be solved by using CV_MAJOR_VERSION instead of OPENCV_VERSION