ROS Conversation Centroid Detection

The project implementation in Master branch covers the requirements for the dissertation project. Please check out the experimental branch for the latest updates and development goals.

Table of Contents

• Introduction
• Installation
• Dependencies
• Usage
• Video Demo

Introduction

This ROS package, ros_conversation_centroid, is designed to detect and track humans in a scene using computer vision techniques. It employs RGB-D data to identify humans and their interactions, and subsequently computes the 2D and 3D centroids of conversations.

The primary features include:

1. Human detection and tracking
2. Social interaction classification (e.g., N-shape, Vis-a-vis)
3. 2D and 3D centroid calculation for conversations

Installation

To install this package, clone the repository into your ROS workspace and build the project:

cd ~/catkin_ws/src
git clone https://github.com/kaustubh285/ros_conversation_centroid/
cd ..
catkin_make
source devel/setup.bash

Dependencies

• ROS (Tested on ROS Noetic)
• OpenCV (Tested on version 4.x)
• Python 3.x
• NumPy

Usage

To run the code, simply execute the following command:

rosrun ros_conversation_centroid conversation_centroid_finder.py

Video Demo

A video demonstration of this project is available on YouTube. You can watch it here.