Advanced machine learning methods with applications in optimization and robot vision - SUPORT

Institutional research project

Area

STEM

Duration

48 months

Total funds

104,000.00 EUR

Abstract

Over the past few decades, machine learning has emerged as a key technology shaping advances in science, industry, and everyday life.

Its wide application made it possible to solve complex problems and understand complex data structures, making it the basis of modern research projects.

This project brings together four complementary research topics with the aim of making a scientific contribution in the field of application of machine learning in real-world data processing:

  1. three-dimensional scene reconstruction in the form of a set of primitives from one RGB-D image,
  2. three-dimensional reconstruction of trees from multiple RGB-D images,
  3. application of machine learning in evolutionary computing for global optimization,
  4. explainable artificial intelligence (Explainable AI – XAI) for identification animal.

The common framework of these topics is the application of machine learning as a methodological approach, providing a powerful tool for data analysis and decision-making in different domains.

The goal of the project is the development of new or improvement of existing methods based on machine learning with a focus on the 4 above-mentioned areas of application.

Main objectives

C1

Increasing the number of published scientific papers by FERIT

C2

Creating new open source computer programs

C3

Creation of publicly available data sets: a data set for three-dimensional scene reconstruction from a single RGB-D image and a data set for creating 3D tree models

Additional goals

Additional project objectives related to the strategic objectives of the financing program:

C4: Defense of one doctoral thesis. (specific objective 1.4)

C5: Involvement of students in scientific work (specific objective 3.1)

C6: Popularization of robotics (specific objective 4.6)

Additional goals of the project – scientific contributions of the proposed research:

C7: Improve existing or develop new methods of three-dimensional modeling of objects and space with regard to applications in robotics. The goal is to develop methods for creating three-dimensional models of the robot’s working environment that will create models with as few parameters as possible with sufficient precision in the shape description.

C8: Develop new or improve existing methods for the reconstruction of 3D wood models based on 3D wood images from multiple views.

C9: Develop new or improve existing methods for verification of reconstructed 3D wood model.

C10: Develop and improve machine learning algorithms to support evolutionary computing algorithms in terms of increasing their efficiency when solving various global optimization problems.

C11: Improve the effectiveness of methods for detection, identification and monitoring of animals with application in livestock.

Research team

prof. dr. sc. Robert Cupec

Project manager and research leader in the field of 3D modeling

doc. dr. sc. Dražen Bajer

Research leader in the field of machine learning applications

doc. dr. sc. Bruno Zorić

Researcher in the field of machine learning applications

doc. dr. sc. Petra Pejić

Research leader in the field of three-dimensional tree reconstruction

dr. sc. Mario Dudjak

Researcher in the field of machine learning applications

dr. sc. Matej Džijan

Researcher in the field of 3D model making

Marina Peko, dipl. ing. el.

Researcher in the field of machine learning applications

Jana Dukić, univ. mag. ing. comp.

Researcher in the field of creating 3D tree models

Funded by the European Union - Next Generation EU. However, the views and opinions expressed are solely those of the author(s) and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.