Repository files navigation

This is the software bundle "gSLICr", a library for real-time superpixel segmentation written in C++ and CUDA. The current version is maintained by:

Carl Yuheng Ren :carl@robots.ox.ac.uk
Victor Adrian Prisacariu :victor@robots.ox.ac.uk
Ian D Reid :ian.reid@adelaide.edu.au

For more information about gSLICr please visit the project websitehttp://www.robots.ox.ac.uk/~victor/gslicr.

Other related projects can be found in the Oxford Active Vision Libraryhttp://www.oxvisionlib.org.

• Works for any size / number of super pixels
• With GTX Titan X, 3.4ms@640x480, 12ms@1280x960, 20ms@1920x1080 image
• Multi-platform supported
  • Win8 Visual Studio
  • Ubuntu 14.04
  • Mac OSX 10.10

Several 3rd party libraries are needed for compiling gSLICr. The given version numbers are checked and working, but different versions might be fine as well. Some of the libraries are optional, and skipping them will reduce functionality.

• cmake (e.g. version 2.8.10.2 or 3.2.3)
  REQUIRED for Windows and Linux, unless you write your own build system
  available athttp://www.cmake.org/

• CUDA (e.g. version 6.0 or 7.0)
  REQUIRED for all GPU accelerated code
  available athttps://developer.nvidia.com/cuda-downloads

• OpenCV (e.g. version 3.0)
  REQUIRED if you want to run the demo, where it is used for reading camera input and displaying UIavailable athttp://opencv.org/downloads.html

To compile the system, use the standard cmake approach:

  $ mkdir build  $ cd build  $ cmake /path/to/gSLICr  $ make

To run demo:

• plug in a webcam
• use the standard cmake approach:

mkdir buildcd buildcmake ../make./demo

• presss to save current segmentation result to current folder.

If you use this code for your research, please kindly cite:

@article{gSLICr_2015,author = {{Ren}, C.~Y and {Prisacariu}, V.~A. and {Reid}, I.~D},title = "{gSLICr: SLIC superpixels at over 250Hz}",journal = {ArXiv e-prints},eprint = {1509.04232},year = 2015,month = sep}

arxiv linke