LBP implementation in multiple computing platforms
Local Binary Pattern (LBP) is a texture operator that is used in several different computer vision applications and implemented in a variety of platforms. When selecting a suitable LBP implementation platform, the specific application and its requirements in terms of performance, size, energy efficiency, cost and developing time has to be carefully considered.

This is a software toolbox that collects the software implementations of the Local Binary Pattern Operator in several platforms:
  • OpenCL for CPU & GPU
  • OpenCL for GPU (branchless)
  • C code optimized for ARM
  • OpenGL ES 2.0 shader language for mobile GPUs
  • C code for TI C64x DSP core (branchless)
  • C code for TTA processor synthesis

If you are using this software in any project, please cite the following article:


Download source code packagelink

from Sourceforge


Thanks to Jani Boutellier, Jari Hannuksela, Henri Nykänen & Sami Varjo
BSIF implementation in Matlab using GPU
Binarized Statistical Image Features (BSIF) is a texture operator that is used in several different computer vision applications. This is a software toolbox extends the original BSIF code allowing the utilization of a GPU in Matlab to compute the features. It contains:
  • Matlab function to calculate BSIF in CPU
  • Matlab function extension to calculate BSIF in GPU
  • Pre-learnt filters
  • Usage instructions

If you are using this software in any research or project, please cite one of the following articles:

Download source code packagelink

Simple classification method for KinFaceW Data Sets
The Kinship Face in the Wild data sets ( KinFaceW I & II ) are commonly used for the evaluation of kinship verification algorithms. We note that the images in the data sets have relationship pairs that in many cases belong to the same original images. This is a simple classification method that utilizes the distance of the chrominance averages of the images in the Lab color space, to classify a pair of images into kinship category. The method achieves 70% recognition accuracy in the KinFaceW-I data set and 80% accuracy in the KinFaceW-II data set, without any training.
If you are using this software in any research or project, please cite one of the following articles:

Download source code packagelink


Contact

Office: TS301


Address:

Department of Computer Science and Engineering
P.O. Box 4500
90014 University of Oulu
Finland


Tel. +358 294 482523
Fax. +358 8 553 2612
email: miguel.bordallo[@AT]ee.oulu.fi