MeanShift

Parallel computing project using Java and Java Threads

Prerequisites

• Java 8
• Maven 3.8.2

Brief introduction

For this project we implemented the Mean Shift, a non-parametric clustering algorithm based on kernel-density estimation that has a quadratic computational complexity.

We have exploited the embarrassingly parallel structure of the algorithm to improve its performance using Java and its parallel mechanisms.

We tried to efficiently use the cache using a certain data organization in memory called Structure of Array and we showed the difference of using the Runnable and Callable interfaces to implement the threads.

To perform the experiments the mean shift clustering has been used to perform image processing, more precisely image segmentation.

For more details on the development process of the project and the tools used see the Final Report.

Experiments setup

To obtain the results shown in the Final Report has been used a machine with:

• Intel© Core™ i7-8750H CPU @ 2.20GHz with 6 cores / 12 threads

Getting started

Using IDE

• Clone this repository in your IDE. Eclipse has been used for this project.
• Set flags in the header of "App.java" to select the implementation of the algorithm that will be run.
• To run the Maven build
  • Run the root directory of the project as Maven build with goals: clean verify
• To run the Java application
  • In your IDE open the file "App.java" and run it as Java application.