CudaVisionSysDeploy

GPU accelerated pedestrian detector using a Pyramidal-Sliding Window approach with HOGLBP as feature descriptor and SVM as the classifier

Software developed by Víctor Campmany (vcampmany@cvc.uab.es) Computer Vision Center and Computer Architecture Department, Universitat Autonoma de Barcelona

How to compile it

Add your own library paths to the Makefile (instruction in the Makefile)

cd into_repository_path
make all

How to run it

./bin/visionSys

Edit the parameters.ini file in order to modify the parameters as desired

Requirements

  • OpenCV >= 2.4
  • CUDA toolkit
  • NVIDIA GPU >= Kepler 3.0

Related publications

GPU-based Pedestrian Detection for Autonomous Driving. V. Campmany, S. Silva, A. Espinosa, JC. Moure, D. Vazquez, A. Lopez. ICCS 2016, Procedia Computer Science 80, 2377-2381