INRIA目标检测和定位工具包 INRIA Object Detection and Localization Toolkit

INRIA Object Detection and Localization Toolkit

---

Author: Navneet Dalal

OLT toolkit for Windows: Wilson Suryajaya Leoputra, Curtin University, Australia, has modified OLT toolkit for windows. You can download the source code from hiswebsite. 

Download the binaries or the library version of the software for Linux from here. Release date: 13 Aug, 2007. NOTE the code accepts only linear SVM models. 
These are OLD binaries. Users are requested to use the code to compile binaries for their system. For info, these binaries are complied and tested on Ubuntu 7.04 using gcc 4.1. The library accepts the same linear SVM model files as dumped by the above binaries. 

Download the binaries of the software for Mac Leopard 10.5.x from here. Release date: 05 Sept, 2008. NOTE the code accepts only linear SVM models. 
Binaries required imlib2 library. You can get it from MacPort and imlib2 contained in there. 

To download the source code, click here. Release date: 16 Oct, 2008 (Uses improved m4 macros, removed GSL dependency, depends upon imlib2, boost 1.35+ and blitz. Compiling should be breeze now. Also added code for compiling a binary which can read binary annotations and dump it in PASCAL format.). NOTE the code accepts only linear SVM models. 


The research work is described in our CVPR 2005 paper Histograms of Oriented Gradients for Human Detection and my PhD thesis Finding People in Images and Vidoes. Besides person detection we have used the same framework for other object detection tasks. You can find the details in my PhD thesis. For results on Pascal year 2005/2006 see Pascal Visual Object Classes Challenge. If you are interested in comparison of our approach to other approaches see results of the Pascal Challenge.

Notes

• The binaries and the source provided above are part of a much bigger toolkit that I have written during my PhD thesis. We have done extensive testing and have evaluated several other approaches for object detection, so some options or pieces of the source code may be redundant for object detection task described in CVPR 2005 and the Pascal Challenge. You may play with these options, but do not send me mails if something stop working if you change options.
• The binary package includes an example shell script and a readme. Read readme, download our associated person detection data set and run the provided example script.
• The code is provided as it is without any support.
• Current documentation is far from sufficient. Please read our CVPR paper and my PhD thesis carefully. Hopefully meaning of most options should be clear.
• For purpose of our experiments, we modified SVM light of Thorsten Joachims. You can download the modified version of SVM light from here. We call this version "SVM dense". Note SVM dense is an unsupported code. Please DO NOT send any mail w.r.t. it neither to Thorsten Joachims nor to us.

---

Disclaimer

THIS CODE IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. Use at your own risk. 

License

This software is distributed under the "CeCILL-B FREE SOFTWARE LICENSE AGREEMENT", and subject to the following additional requirement (Article 14) regarding acknowledgements.

ARTICLE 14 - ACKNOWLEDGEMENTS

Any Licensee who publishes a scientific or technical Work (including but not limited to conference and journal papers, technical reports, white papers, software manuals, web sites, and oral presentations of these) that is based on, or that uses results obtained with, the original or Modified Software hereby expressly agrees to include in the Work an acknowledgement of the use of the Software in the Work. Moreover, if the Work includes scientific citations, and unless such a citation would generally be considered to be inappropriate, the Licensee agrees to include in the Work at least one citation to the scientific publications that originally described the method implemented in the Software. 

The following scientific publications contain the original descriptions of the method implemented in this Software:

• Histograms of Oriented Gradients for Human Detection. Navneet Dalal and Bill Triggs. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), vol. II, pages 886-893, June 2005. Paper available here.
• Human detection using oriented histograms of flow and appearance. Navneet Dalal, Bill Triggs and Cordelia Schmid. European Conference on Computer Vision (ECCV), vol. II, pages 428-441, May 2006. Paper available here.
• Finding people in images and videos. Navneet Dalal. PhD thesis, Institut National Polytechnique de Grenoble, July 2006. Thesis available here.

---

Support of European Union 6th framework project aceMedia is greatly acknowledged.

---

Serbo-Croatian translation of this page.

from: http://pascal.inrialpes.fr/soft/olt/