Projects
Search in art image collections based on color semantics Hasselt University
Prevention and remediation of soil compaction Research Institute for Agriculture, Fisheries and Food
At the start of the 4-year VLAIO research project "Preventing and Remedying Soil Compaction" in 2019, soil compaction was still only marginally on the radar of farmers, contractors and processors. However, soil compaction affects about one-third of the arable land in Belgium; indeed, across Europe, the phenomenon is considered a serious threat to agricultural production. The use of increasingly heavier ...
Efficient generation and compression of digital holograms with deep neural network encoding. Vrije Universiteit Brussel
Data Visualisation and Visual Design in Biological and Agricultural Sciences: Unearthing Complex Insights KU Leuven
The project 'Diving Deep into Biological Data Complexity' seeks to address these challenges through the application of visual analytics and topological data analysis. Visual analytics offers a human-centric approach to data exploration, allowing researchers to intuitively interact with their contextualised data and draw insights that may be hidden in purely statistical analyses. In addition, topological data analysis will be used to capture ...
Multimodal Hopfield networks: a step towards next-generation AI Ghent University
The field of AI is in constant need for more scalable solutions. One promising option are Hopfield networks, that allow for an analog electronics implementation that is faster and much more energy-efficient than modern digital hardware. Unlike traditional feedforward neural networks, these networks also incorporate feedback connections, inspired by the structure of the human brain. Nonetheless, current models are typically limited to unimodal ...
Multimodal Hopfield networks: a step towards next-generation AI Ghent University
The field of AI is in constant need for more scalable solutions. One promising option are Hopfield networks, that allow for an analog electronics implementation that is faster and much more energy efficient than modern digital hardware. Unlike traditional feedforward neural networks, these networks also incorporate feedback connections, inspired by the structure of the human brain. Nonetheless, current models are typically limited to unimodal ...
Open-set object recognition KU Leuven
Object recognition typically involves classifying or detecting objects based on a pre-defined set of classes. This is known as the closed set assumption: the same set of classes is assumed to be present in both the training and testing data. In contrast, in open-set recognition the goal is to train models that can cope with previously unseen object categories. In this project, we will explore techniques for automatically adapting models to ...
Breaking the Resolution Limit in Two-Photon Microscopy Using Negative Photochromism KU Leuven
Multiphoton microscopy is a benchmark tool in biomedical research, used for the fluorescence imaging in cellular environments. This has important implications for disease diagnosis and the monitoring of therapy response. In conventional two-photon microscopy the fluorescence intensity of the employed molecular probe is proportional to the square of the excitation light intensity, implying that the fluorescence from the sample is confined ...