NOTE: I have moved to the Vision Group at Queen Mary University of London. My new page is located HERE.

Fast Bayesian Object Detection

Reference implementation of our object detector [1]. You will need the following software:

  • OpenCV 2+ (tested with 2.3.2 and 2.4.0)
  • CMake 2.8+ (tested with 2.8)
  • C++ compiler (tested with GCC 4.7)
  • OpenMP (optional but recommended)
  • boost (optional, for the demonstration binary only)

The ZIP archive includes documentation and minimal usage examples. Tested on Debian 7. Not tested under Windows, but it should run just fine. Please contact me if you run into trouble.

Download ZIP Archive - Download Sample Dataset

[1] Kasim Terzić and J.M.H. du Buf: An Efficient Naive Bayes Approach to Category-Level Object Detection, Proc. ICIP 2014, Paris, Oct 2014

Biologically-Inspired Multiscale Keypoints (BIMP)

Optimized CPU-based implementation of our keypoint detector [2,3]. You will need the following software:

  • OpenCV 2+ (tested with 2.3.2 and 2.4.0)
  • CMake 2.8+ (tested with 2.8)
  • C++ compiler (tested with GCC 4.7, 4.8 and MinGW)
  • OpenMP (optional but recommended)
  • Doxygen (for documentation, optional but recommended)

The ZIP archive includes documentation and minimal usage examples. Tested on Debian 6 and 7, Ubuntu 12.10, Linux Mint 13, Windows 7 (with MinGW). Please contact me if you run into trouble.

Download ZIP Archive

[2] Kasim Terzić, J.M.F. Rodrigues and J.M.H. du Buf: Fast Cortical Keypoints for Real-Time Object Recognition, Proc. ICIP 2013, Melbourne, Sep 2013

[3] Kasim Terzić, J.M.F. Rodrigues and J.M.H. du Buf: BIMP: A Real-Time Biological Model of Multi-Scale Keypoint Detection in V1, Neurocomputing, Oct 2014, in press

CUDA-BIMP

CUDA-based GPU implementation of the same detector. This version provides similar results as the reference implementation, but runs in real time. You will need the following software:

  • OpenCV 2+ (tested with 2.3.2 and 2.4.0)
  • CMake 2.8+ (tested with 2.8)
  • C++ compiler (tested with GCC 4.4)
  • CUDA 4.1 + GPU SDK + Nvidia binary drivers
  • OpenMP (optional but recommended)
  • boost::thread (optional but recommended)

Tested on an Nvidia GTX 560 Ti and Debian 7.

Download ZIP Archive

OpenCL-BIMP

OpenCL-based GPU implementation of the same detector. This version provides similar results as the reference implementation, but runs in real time. You will need the following software:

  • OpenCV 2+ (tested with 2.3.2 and 2.4.0)
  • CMake 2.8+ (tested with 2.8)
  • C++ compiler (tested with GCC 4.7)
  • OpenCL driver, libraries, and headers
  • OpenMP (optional but recommended)
  • boost::thread (optional but recommended)

Tested on an Nvidia GTX 560 Ti and Debian 7.

Download ZIP Archive