Section: Software

Histogram of oriented gradient object detection

Participants : Navneet Dalal, Bill Triggs, Cordelia Schmid.

As part of the European Union FP6 Integrated Project aceMedia we have developed a toolkit for detecting specific visual object classes such as humans, cars and motorbikes in static images. Although developed originally for human detection [4] , the software implements a generic framework that can be trained to detect any visual class with a moderately stable appearance. The method has proven quite popular owing to its accuracy and its relative simplicity, with at least six academic or corporate research groups independently reimplementing it and more than 60 first-time downloads (http://lear.inrialpes.fr/software ) since September 2005. The software is under copyright protection, registered at the Agence pour la Protection des Programmes (APP).