AI_Object_Detection

This repository contains all the software regarding our Convolutional Neural Networks implementation on the Jetson Nano. We actually implemented two Convolutional Neural Networks:

1. multiped: it was already trained and detects humans and their luggages
2. a custom Mobilenet-SSD: we trained it ourselves to fit specific classes:
   • person
   • baggage
   • ball
   • bicycle
   • bus
   • car
   • cat
   • dog
   • motorcycle

This software basically detects humans or objects of the previously listed classes and sends video streaming with the detections on an image ROS topic and UInt8 ROS messages on a ROS topic named /detection that corresponds to one of the classes (see classes corresponding numbers).

A more detailled README can be found in the folder where all the scripts are located: scripts folder

How to go to the script repository

Execute cd ./aarch64/bin/GEIcar_Project in a terminal opened in this folder.

How to use the launch files

• The first launch file is designed to change the ROS_IP and the ROS_MASTER_URI as the roscore is not running on the Jetson Nano but on the Raspberry PI. It then runs the detection main script. If it is your case execute the command ./launch.sh in any terminal opened in this folder.

  NB: DO NOT FORGET TO MODIFY ROS_IP AND ROS_MASTER_URI TO FIT YOUR ROS CONFIGURATION.

• If you want to test the software with a local roscore, then execute the command ./launch_local.sh in any terminal opened in this folder after having runned a roscore.