A5043511500- Sparse and Compressive Sensing Techniques
- Wireless Communication Security Techniques
- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Advanced Data Compression Techniques
- Cooperative Communication and Network Coding
- Advanced Neural Network Applications
- Topic Modeling
- Image Processing Techniques and Applications
- Blind Source Separation Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Chaos-based Image/Signal Encryption
- COVID-19 diagnosis using AI
- Air Quality Monitoring and Forecasting
- Explainable Artificial Intelligence (XAI)
- Air Quality and Health Impacts
- Photoacoustic and Ultrasonic Imaging
- Indoor and Outdoor Localization Technologies
- Advanced Vision and Imaging
- Video Coding and Compression Technologies
- Microwave Imaging and Scattering Analysis
- Error Correcting Code Techniques
- Anomaly Detection Techniques and Applications
- Advanced Graph Neural Networks
Vrije Universiteit Brussel
2016-2025
IMEC
2017-2025
Weizmann Institute of Science
2024
Athens University of Economics and Business
2023
Ghent University
2022
University of Salerno
2022
University College London
2013-2017
iMinds
2011-2016
University of Patras
2006-2016
Vrije Universiteit Amsterdam
2009-2015
We address the problem of compressed sensing (CS) with prior information: reconstruct a target CS signal aid similar that is known beforehand, our information. integrate additional knowledge into via l <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -l and xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> minimization. then establish bounds on number measurements required by these problems to successfully original signal. Our...
A new convolution neural network (CNN) architecture for semantic segmentation of high resolution aerial imagery is proposed in this paper. The follows an hourglass-shaped (HSN) design being structured into encoding and decoding stages. By taking advantage recent advances CNN designs, we use the composed inception module to replace common convolutional layers, providing with multi-scale receptive areas rich context. Additionally, order reduce spatial ambiguities up-sampling stage, skip...
Time-synchronized channel hopping (TSCH) is currently the most efficient solution for collision-free interference-avoiding communications in ad hoc wireless networks, such as sensor vehicular and networks of robots or drones. However, all variants TSCH require some form centralized coordination to maintain time-frequency slotting mechanism. This leads slow convergence steady state moderate slot utilization, particularly under node churn mobility. We propose decentralized time-synchronized...
Real-world data processing problems often involve various image modalities associated with a certain scene, including RGB images, infrared images or multi-spectral images. The fact that different share attributes, such as edges, textures and other structure primitives, represents an opportunity to enhance tasks. This paper proposes new approach construct high-resolution (HR) version of low-resolution (LR) given another HR modality reference, based on joint sparse representations induced by...
Natural language explanation (NLE) models aim at explaining the decision-making process of a black box system via generating natural sentences which are human-friendly, high-level and fine-grained. Current NLE <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> Throughout this paper, we refer to as Language Explanation aimed for vision vision-language tasks. explain or model (a.k.a., task model), e.g., VQA model, GPT. Other than additional...
The reconstruction of a high resolution image given low observation is an ill-posed inverse problem in imaging. Deep learning methods rely on training data to learn end-to-end mapping from low-resolution input high-resolution output. Unlike existing deep multimodal models that do not incorporate domain knowledge about the problem, we propose design incorporates sparse priors and allows effective integration information another modality into network architecture. Our solution relies novel...
Internet-of-Things (IoT) technologies incorporate a large number of different sensing devices and communication to collect amount data for various applications. Smart cities employ IoT infrastructures build services useful the administration city citizens. In this article, we present an pipeline acquisition, processing, visualization air pollution over Antwerp, Belgium. Our system employs mounted on vehicles as well static reference stations measure variety parameters, such humidity,...
Training deep neural networks on large datasets containing high-dimensional data requires a amount of computation. A solution to this problem is data-parallel distributed training, where model replicated into several computational nodes that have access different chunks the data. This approach, however, entails high communication rates and latency because computed gradients need be shared among at every iteration. The becomes more pronounced in case there wireless between (i.e., due limited...
Recent success in the field of single image super-resolution (SISR) is achieved by optimizing deep convolutional neural networks (CNNs) space with L1 or L2 loss. However, when trained these loss functions, models usually fail to recover sharp edges present high-resolution (HR) images for reason that model tends give a statistical average potential HR solutions. During our research, we observe gradient maps generated have significantly lower variance than original images. In this work,...
We address the problem of Compressed Sensing (CS) with side information. Namely, when reconstructing a target CS signal, we assume access to similar signal. This additional knowledge, information, is integrated into via ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -ℓ and xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> minimization. then provide lower bounds on number measurements that these problems require for successful...
We propose and analyze an online algorithm for reconstructing a sequence of signals from limited number linear measurements.The are assumed sparse, with unknown support, evolve over time according to generic nonlinear dynamical model.Our algorithm, based on recent theoretical results ℓ1-ℓ1 minimization, is recursive computes the measurements be taken at each on-thefly.As example, we apply compressive video background subtraction, problem that can stated as follows: given set images static...
We address the problem of compressed sensing (CS) with prior information: reconstruct a target CS signal aid similar that is known beforehand, our information. integrate additional knowledge into via L1-L1 and L1-L2 minimization. then establish bounds on number measurements required by these problems to successfully original signal. Our geometrical interpretations reveal if information has good enough quality, minimization improves performance dramatically. In contrast, very classical brings...
Challenges drive the state-of-the-art of automated medical image analysis. The quantity public training data that they provide can limit performance their solutions. Public access to methodology for these solutions remains absent. This study implements Type Three (T3) challenge format, which allows on private and guarantees reusable methodologies. With T3, organizers train a codebase provided by participants sequestered data. T3 was implemented in STOIC2021 challenge, with goal predicting...
In the context of low-cost video encoding, distributed coding (DVC) has recently emerged as a potential candidate for uplink-oriented applications. This paper builds on concept correlation channel (CC) modeling, which expresses noise being statistically dependent side information (SI). Compared with classical side-information-independent (SII) modeling adopted in current DVC solutions, it is theoretically proven that side-information-dependent (SID) improves Wyner-Ziv performance. Anchored...
The problem of predicting the location users on large social networks like Twitter has emerged from real-life applications such as unrest detection and online marketing. user geolocation is a difficult active research topic with vast literature. Most proposed methods follow either content-based or network-based approach. former exploits user-generated content while latter utilizes connection interaction between users. In this paper, we introduce novel method combining strength both...
In support of art investigation, we propose a new source separation method that unmixes single X-ray scan acquired from double-sided paintings. this problem, the signals to be separated have similar morphological characteristics, which brings previous methods their limits. Our solution is use photographs taken front-and back-side panel drive process. The crux our approach relies on coupling two imaging modalities (photographs and X-rays) using novel coupled dictionary learning framework able...
We propose a recursive algorithm for estimating time-varying signals from few linear measurements. The are assumed sparse, with unknown support, and described by dynamical model. In each iteration, the solves an ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -ℓ minimization problem estimates number of measurements that it has to take at next iteration. These computed based on recent theoretical results minimization. also provide...