PeTra (People Tracking)

PeTra is a tool which allows detecting and tracking people. The system is based on a CNN that uses an occupancy map constructed from the readings of a LIDAR sensor.

• Dataset: https://www.frontiersin.org/articles/10.3389/fnbot.2017.00072/full
• PeTra:
  • https://www.frontiersin.org/articles/10.3389/fnbot.2018.00085/full
  • https://www.mdpi.com/2218-6581/8/3/75

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.

Prerequisites

• ROS KINECTIC or ROS MELODIC
• ROS TensorFlow installation

  $ cd ~/catkin_ws/src/  
  && git clone https://github.com/ethz-asl/tensorflow_catkin.git  
  && cd tensorflow_catkin/  
  && git checkout bcd27b06e65dd88c3f2f2875ff498b3c2bddcff3  
  && cd ../  
  & git clone https://github.com/catkin/catkin_simple.git  
  && cd catkin_simple/  
  && git checkout 0e62848b12da76c8cc58a1add42b4f894d1ac21e  

  To compile tensorflow_catkin package with CPU optimization. Open the CMakeLists.txt file and change to ON the flag: "-Dtensorflow_OPTIMIZE_FOR_NATIVE_ARCH=OFF"

$ cd ~/catkin_ws/
$ catkin_make  
$ cd ~/catkin_ws/src/
$ git clone https://github.com/tradr-project/tensorflow_ros_cpp.git
$ cd ..
$ catkin_make

PeTra Installation

 $ cd ~/catkin_ws/src/  
 $ git clone https://github.com/ClaudiaAlvarezAparicio/petra.git
 $ cd ..  
 $ catkin_make  

Configuration

Edit the different parameters of petra/config/parameters.yaml

Execution

Execution with a rosbag:

roslaunch petra petra_rosbag.launch rosbag_file:=absolute_path_to_bag_file

Execution in real time:

roslaunch petra petra.launch

Docker Image

To test PeTra we have create tow docker images, steps to test it:

• ROS Kinetic

  $ docker pull claudiaalvarezaparicio/petra:kinetic  
  $ docker run -d -p 6901:6901 -e VNC_PW=petra --name=petra claudiaalvarezaparicio/petra:kinetic  

• ROS Melodic

  $ docker pull claudiaalvarezaparicio/petra:melodic  
  $ docker run -d -p 6901:6901 -e VNC_PW=petra --name=petra claudiaalvarezaparicio/petra:melodic  

In the browser: http://localhost:6901/
Password: petra

Execute PeTra in docker

1.- Open terminal:

$ roscore  

2.- Open terminal:

$ rosrun rviz rviz  

3.- Open terminal

$ roslaunch petra petra_rosbag.launch rosbag_file:=/home/student/rosbags/kitchen1.bag  

Stop and Remove Docker Container

$ docker stop petra
$ docker rm petra 

Hardware Specifications

The tool has been test with a Hokuyo URG-04LX-UG01 (https://www.hokuyo-aut.jp/search/single.php?serial=166).

With another LIDAR it can work but is better train a new neural network model.

Data Labelig for Training

To train a new neural network model is necessary to get the LIDAR data and label it. The data_labeling folder has a README to label data with a KIO RTLS system or with the own PeTra.

Network Model Training

Once the data to train the network is ready, its time to train the model. The neural_network folder contains a README with the necessary steps to train the model.