/didi-challenge-17

ROS Implementation of our solution to Didi Challenge 2017.

Primary LanguageC++

Didi Challenge 17

ROS Implementation of our solution to Didi Challenge 2017.

Package developed at Intelligent Systems Laboratory, Universidad Carlos III de Madrid.

Prerequisites

Software

  • Ubuntu 14.04/16.04
  • ROS Indigo/Kinetic
  • OpenCV 3.2 with opencv__contrib installed
  • Velodyne ROS driver from here
  • Python 2.7
  • C++11
  • CUDA

Hardware

  • NVIDIA GPU for Deep Learning Classifier

Setup

  1. Clone this repository into your catkin workspace ;)

  2. Fetch submodule dependencies:

    cd ~/catkin_ws/src/didi-challenge-17
git submodule update --init --recursive

  3. Follow these instructions to install and configure the classifier dependencies (Mainly Caffe and pycaffe)

  4. Compile the ROS nodes

    cd ~/catkin_ws/
catkin_make

How to run?

For cars:

roslaunch didi_challenge awesome_classifier.launch

For pedestrians:

roslaunch didi_challenge pedestrian_classifier.launch

Known issues

If compiler complains about some OpenCv dependency, make sure CMakeLists.txt points to the actual OpenCv installation path in your computer (remember to install them along with opencv__contrib)