Courses

Below you can find the courses I have been involved with during my doctoral and postdoctoral career at Ghent University.

Computer graphics (2014-Present)

The course computer graphics (ECTS-card), lectured by prof. Aleksandra Pizurica, is taught in the Master Computer Science Engineering. I take care of the projects of the course, which consist of a programming assignment (using OpenGL or VTK) and focus on a particular subtopic of computer graphics. Among the topics are 3D scene generation, 3D space partitioning and mesh simplification.

Research project (2020-Present)

Research Project (ECTS-card) is a course in the Master Computer Science Engineering. The aim of the course is to teach the students elementary elements of scientific research, and especially on the process of searching the literature and scientific writing.

2021 - Multi-agent SLAM

Simultaneous Localization And Mapping (SLAM) is the problem of giving a robot (or agent) placed at an unknown location in an unknown environment the ability to incrementally build a consistent map of this environment while simultaneously determining its location within this map [Durrant-Whyte2006]. Historically, SLAM methods have been designed mostly for single-agent scenarios: the agent builds a map and localizes itself as it evolves in the area, without any information exchange with peers [Mur-Artal2016]. For a wide range of use cases, extending these systems to allow multi-agent operation would bring a lot of value: the area to be mapped can be covered more quickly and information can be shared, reused and corrected by the different agents, leading to a potential improvement in speed and accuracy.

In a multi-agent SLAM system, such as the one explained in the paper ‘Multi-robot collaborative dense scene reconstruction’ by Dong et al., several rovers collaborate to build a single 3D map of the environment

2020 - Hyperspectral image classification

The goal of hyperspectral image classification is to assign a class label to each pixel in the hyperspectral image. Improvements in spectral resolution have lead to advances in signal processing and machine learning algorithms. However, the classification of hyperspectral images still presents many challenges. If compared with traditional RGB images, the volume of hyperspectral data might be similar but the structure is completely different.

Most hyperspectral sensors have a low spatial resolution so deep learning techniques designed for computer vision are not easily transferable to hyperspectral data cubes since the spectral dimension prevails over the spatial neighborhood in most cases. Moreover, the low spatial resolution limits the number of samples available for training statistical models. This also makes the annotation process - which is required in supervised learning - more difficult since objects smaller than the spatial resolution are mixed with their neighborhood.

Apart from the low spatial resolution, other challenges related to the hyperspectral image classification are the presence of redundant features, the imbalance among the limited number of available training samples, and the high dimensionality of the data. All this means that HSI classification remains a highly active field in compute science research.

You can find the final survey paper of the students on this topic [here]. The goal of hyperspectral image classification is to assign a class label to each pixel in the image. An RGB-image (a) with its corresponding HSI cube (b) and the output of the classification (c).

Bachelor project (2020)

Bachelor Project is a course in the bachelor Industrial Engineering: Electronics ICT (ECTS-card). The students use the skills and knowledge from different courses and apply them to an existing problem that is relevant for today’s industry. I supervised a student who has worked on a novel visualization tool for hyperspectral image processing that allows efficient querying in both the temporal, spatial and spectral dimension.

Screenshot from the hyperspectral imaging visualization tool, integrating efficient querying in the temporal (below), spatial (top left) and spectral (top right) dimension, currently unavailable in the HSI community.

Information technology and data processing (2014-2021)

The course information technology and data processing (ECTS-card), lectured by prof. Sofie Verbrugge and dr. Jan Aelterman, is taught in the Master of Science in Industrial Engineering and Operations Research. I assist the practical exercise sessions, where students are taught on how to query a relational database by means of the SQL query language.