A5042628612- Metaheuristic Optimization Algorithms Research
- Evolutionary Algorithms and Applications
- Neural Networks and Applications
- Advanced Multi-Objective Optimization Algorithms
- Generative Adversarial Networks and Image Synthesis
- Machine Learning and Data Classification
- Remote Sensing and LiDAR Applications
- Face and Expression Recognition
- Remote Sensing in Agriculture
- Radiomics and Machine Learning in Medical Imaging
- Machine Learning and Algorithms
- Domain Adaptation and Few-Shot Learning
- Osteoarthritis Treatment and Mechanisms
- Advanced Neural Network Applications
- Gaussian Processes and Bayesian Inference
- Advanced Image and Video Retrieval Techniques
- Reinforcement Learning in Robotics
- Forest ecology and management
- Medical Image Segmentation Techniques
- Blind Source Separation Techniques
- Algorithms and Data Compression
- Speech Recognition and Synthesis
- Model Reduction and Neural Networks
- COVID-19 diagnosis using AI
- Conservation, Biodiversity, and Resource Management
University of Copenhagen
2016-2025
IT University of Copenhagen
2024
Laboratoire des Sciences du Climat et de l'Environnement
2023
University of Rwanda
2023
California Institute of Technology
2023
German Research Centre for Artificial Intelligence
2023
Deutsches Forschungsnetz
2023
Fraunhofer Institute for Digital Medicine
2022
TU Dortmund University
1997-2011
Ruhr University Bochum
2001-2010
The “German Traffic Sign Recognition Benchmark” is a multi-category classification competition held at IJCNN 2011. Automatic recognition of traffic signs required in advanced driver assistance systems and constitutes challenging real-world computer vision pattern problem. A comprehensive, lifelike dataset more than 50,000 sign images has been collected. It reflects the strong variations visual appearance due to distance, illumination, weather conditions, partial occlusions, rotations. are...
Real-time detection of traffic signs, the task pinpointing a sign's location in natural images, is challenging computer vision high industrial relevance. Various algorithms have been proposed, and advanced driver assistance systems supporting recognition signs reached market. Despite many competing approaches, there no clear consensus on what state-of-the-art this field is. This can be accounted to lack comprehensive, unbiased comparisons those methods. We aim at closing gap by "German...
The covariancematrix adaptation evolution strategy (CMA-ES) is one of themost powerful evolutionary algorithms for real-valued single-objective optimization. In this paper, we develop a variant the CMA-ES multi-objective optimization (MOO). We first introduce single-objective, elitist using plus-selection and step size control based on success rule. This algorithm compared to standard CMA-ES. turns out be slightly faster unimodal functions, but more prone getting stuck in sub-optimal local...
In this work, we report the set-up and results of Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with IEEE International Symposium on Biomedical Imaging (ISBI) 2017 Conferences Medical Image Computing Computer-Assisted Intervention (MICCAI) 2018. The image dataset is diverse contains primary secondary tumors varied sizes appearances various lesion-to-background levels (hyper-/hypo-dense), created collaboration seven hospitals research institutions. Seventy-five...
Mammographic risk scoring has commonly been automated by extracting a set of handcrafted features from mammograms, and relating the responses directly or indirectly to breast cancer risk. We present method that learns feature hierarchy unlabeled data. When learned are used as input simple classifier, two different tasks can be addressed: i) density segmentation, ii) mammographic texture. The proposed model at multiple scales. To control models capacity novel sparsity regularizer is...
In this work, we report the set-up and results of Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with IEEE International Symposium on Biomedical Imaging (ISBI) 2017 Conferences Medical Image Computing Computer-Assisted Intervention (MICCAI) 2018. The image dataset is diverse contains primary secondary tumors varied sizes appearances various lesion-to-background levels (hyper-/hypo-dense), created collaboration seven hospitals research institutions. Seventy-five...
Sleep disorders affect a large portion of the global population and are strong predictors morbidity all-cause mortality. staging segments period sleep into sequence phases providing basis for most clinical decisions in medicine. Manual is difficult time-consuming as experts must evaluate hours polysomnography (PSG) recordings with electroencephalography (EEG) electrooculography (EOG) data each patient. Here, we present U-Sleep, publicly available, ready-to-use deep-learning-based system...
Abstract Cognitive impairment in patients with Alzheimer's disease (AD) is associated reduction hippocampal volume magnetic resonance imaging (MRI). However, it unknown whether texture changes persons mild cognitive (MCI) that does not have a change volume. We tested the hypothesis has association to early loss beyond of volumetric changes. The marker was trained and evaluated using T1‐weighted MRI scans from Disease Neuroimaging Initiative (ADNI) database, subsequently applied score...
Abstract Trees sustain livelihoods and mitigate climate change but a predominance of trees outside forests limited resources make it difficult for many tropical countries to conduct automated nation-wide inventories. Here, we propose an approach map the carbon stock each individual overstory tree at national scale Rwanda using aerial imagery from 2008 deep learning. We show that 72% mapped are located in farmlands savannas 17% plantations, accounting 48.6% aboveground stocks. Natural cover...
The distribution of dryland trees and their density, cover, size, mass carbon content are not well known at sub-continental to continental scales1-14. This information is important for ecological protection, accounting, climate mitigation restoration efforts ecosystems15-18. We assessed more than 9.9 billion derived from 300,000 satellite images, covering semi-arid sub-Saharan Africa north the Equator. attributed wood, foliage root every tree in 0-1,000 mm year-1 rainfall zone by coupling...
Abstract The consistent monitoring of trees both inside and outside forests is key to sustainable land management. Current systems either ignore or are too expensive be applied consistently across countries on a repeated basis. Here we use the PlanetScope nanosatellite constellation, which delivers global very high-resolution daily imagery, map forest non-forest tree cover for continental Africa using images from single year. Our prototype 2019 (RMSE = 9.57%, bias −6.9%). demonstrates that...
Trees are an integral part in European landscapes, but only forest resources systematically assessed by national inventories. The contribution of urban and agricultural trees to national-level carbon stocks remains largely unknown. Here we produced canopy cover, height above-ground biomass maps from 3-meter resolution nanosatellite imagery across Europe. Our estimates have a systematic bias 7.6% (overestimation;
Quantifying forest biomass stocks and their dynamics is important for implementing effective climate change mitigation measures by aiding local management, studying processes driving af-, re-, deforestation, improving the accuracy of carbon accounting. Owing to 3-dimensional nature structure, remote sensing using airborne LiDAR can be used perform these measurements vegetation structure at large scale. Harnessing full dimensionality data, we present deep learning systems predicting wood...
First, the covariance matrix adaptation (CMA) with rank-one update is introduced into (1+1)-evolution strategy. An improved implementation of 1/5-th success rule proposed for step size adaptation, which replaces cumulative path length control. Second, an incremental Cholesky developed replacing computational demanding and numerically involved decomposition matrix. The can replace only without evolution reduces effort from O(n3) to O(n2). resulting (1+1)-Cholesky-CMA-ES elegant algorithm...
This paper presents a brain T1-weighted structural magnetic resonance imaging (MRI) biomarker that combines several individual MRI biomarkers (cortical thickness measurements, volumetric hippocampal shape, and texture). The method was developed, trained, evaluated using two publicly available reference datasets: standardized dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) arm of Australian Imaging Biomarkers Lifestyle flagship study ageing (AIBL). In addition, by...
In machine learning, active learning refers to algorithms that autonomously select the data points from which they will learn. There are many mining applications in large amounts of unlabeled readily available, but labels (e.g., human annotations or results coming complex experiments) costly obtain. such scenarios, an algorithm aims at identifying that, if labeled and used for training, would most improve learned model. Labels then obtained only promising points. This speeds up reduces...