Welcome to the Computer Vision and Machine Learning Systems Group
We are interested in interdisciplinary research questions involving the development of novel computer vision, machine learning and imaging technologies yielding new approaches to acquire and analyse data with applications in all kinds of disciplines including ecology, medicine, biology, additive manufacturing, robotics and digital humanities. This requires a fundamental investigation of how computers perceive and understand complex real-world situations. By examining the entire process from data acquisition (i.e. sensing hardware) over data interaction (e.g. augmented and virtual reality) to quantitative evaluations (i.e. algorithms) we are seeking for solutions beyond classical image analysis, pattern recognition and artificial intelligence methodologies. Many lessons remain to be learned to tackle the challenges of real-world data which we seek to study and reveal in the coming years to develop novel and sustainable data-driven systems. If you would like to get in touch with us please contact Prof. Dr. Benjamin Risse.
Our joint project "maQinto: Machine-trained quality sensor, intelligent process control and an ML framework for resource-efficient, customized carbon fiber production" is funded by the Bundesministerium für Bildung und Forschung. We are looking forward to work closely in this interdisciplinary consortium.
Our paper "Cell Selection-based Data Reduction Pipeline for Whole Slide Image Analysis of Acute Myeloid Leukemia" has been accepted at the International Conference on Computer Vision and Pattern Recognition (CVPR) at the Computer Vision for Medical Images (CVMI) workshop.
Jacqueline Kockwelp won the Price for Applied Computer Science for her master thesis on AML prediction. Congratulations! The press release can be found here.
Our paper "Narrowing Attention in Capsule Networks" has been accepted and will be presented a the International Conference on Pattern Recognition (ICPR 2022) in Montréal Quèbec. In this paper a new attention-based routing strategy for deep capsule networks is presented.
Our conference abstract on "Deriving trajectories from videos to calibrate and validate agent-based models" has been accepted to the iEMSs 2022. Special thanks to Judith Verstegen and her team.
Our paper "Towards VR Simulation-Based Training in Brain Death Determination" has been accepted at the IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW). In this work we demonstrate our first insights of our VR-based medical training research. The paper can be found here.
The Computer Vision & Machine Leanring Systems group has recently moved to the Institute for Geoinformaitcs at the University of Münster. We are more than happy to be part of this excellent community and we are looking forward to many exciting endevours ahead.
Our project "GIGA Sign Language - Development of a 5G enabled Sign Language Application" has been funded by the Government of Nordrhein-Westfalen. Our group is part of a bigger consortium consisting of academic and industrial partners to implement a Machine Learning-based German sign language App. The official press release by the Government can be found here and the University announcement is given here.
CVMLS | | Paper Accepted in Nature Communications!
The paper "Perspectives in machine learning for wildlife conservation" has been accepted in Nature communications and provides an overview of latest sensing and deep learning technologies to monitor vertebrates accross scales. The paper can be found here.
Our research project on "Neuronal fingerprints as structure- and activity-sensitive molecular representations" has been funded by the Deutsche Forschungsgemeinschaft. The Project is part of the Priority Program SPP2363 Molecular Machine Learning (MML). Special thanks to our collaborator (AG Koch).
Our team is part of a consortium which won the Teaching Award of the University of Münster 2021 for our virtual reality implementation for training medical students. The University press release can be found here and additional informaiton are linked here. A video demonstrating our system can be found on YouTube here (German only).
The paper "An uncertainty-aware, shareable, and transparent neural network architecture for brain-age modeling" has been accepted in Science Adances and can be found here.
CVMLS | | Project Funded by the Government of NRW!
Our REACT EU project on "Computer Vision and Deep Learning for AML Diagnosis" has been funded by the Government of Nordrhein-Westfalen. Special thanks Linus Angenedt and Christoph Schliemann of the University Hospital in Münster.
Our podcast "AI - Friend or Enemy" ("KI - Freund oder Feind") has been released. For some, artificial intelligence (AI) is a beacon of hope, for others a threat to their autonomy. Prof. Dr. Benjamin Risse, who researches in AI, explains in the podcast whether intelligent systems and machines are more likely to help or hinder us in our social development. You can find the podcast here (German only).
We have joined the Health-AI Network. Health.AI taps the potential of artificial intelligence in a broad, multidisciplinary partner network. We are therefore shaping this process together with partners from research, business, politics and civil society. For more information see here.
Our REACH incub.AI.tor podcast has been published. In this episode, Dr. Anne Vortkamp talks to two of the initiators of the so-called incub.AI.tors, which is all about artificial intelligence. Prof. Dr. Benjamin Risse from the Department of Computer Science and Holger Angenent from WWU-IT tell us what the REACH-funded idea is all about and why it can be useful and important for startups to use artificial intelligence. The podcast (German only) can be found here.
Our joint project "medical tr.AI.ning - Intelligent Virtual Agents for Medical Education" has been funded by the Federal Ministry of Education and Research. The project aims to establish an AI-based simulation and training platform which serves to promote clinically oriented logical thinking of future physicians through the use of lifelike virtual reality simulations. The press release of the University can be found here and the announcement of the University Hospital is linked here.
CVMLS | | inFlame Medical Scientists College Funded!
The inFlame Medical Scientists College is funded by the Else Kröner-Fresenius Foundation. This project aims to promote collaboration between basic research and clinical applications. The project of the CVMLS group is particularly interested in using Machine Learning algorithms in biomedical research. The University press release can be found here and the University Hospital announcement is lined here.
The Interdisciplinary Teaching Program on Machine Learning and Artificial Intelligence, in short InterKIWWU, is funded by the funding initiative "Künstliche Intelligenz in der Hochschulbildung" of the Federal Ministry of Education and Research. The overall aim is to establish and test a graded university-wide teaching program on Machine Learning (ML) and Artificial Intelligence (AI). This project is lead by the CeNoS of the Universtiy of Münster. More information can be found here and the press press release of the University can be found here.
The paper "PHOTONAI—A Python API for rapid machine learning model development" has been published in PLOS ONE. Special thanks to Ramona Leenings, Tim Hahn and everybody involved. The paper can be found here. If you are interested in PHOTONAI you can find the website here.
Our research on Virtual Reality and Machine Learning for medical education has been featured in the news. The WN newspaper article can be found here and the universisty news post is linked here.
Our collaborative research project on developing novel AI algorithms to study plant pollinator interactions has been funded by the Human Frontier Science Program (HFSP). In particular we are interested in gaining insights into more sustainable and efficient agricultural food production methods. Especially in times of global warming and insect defaunation, these interactions will play an important role in the future. More information can be found here.
Our paper "Touch Recognition on Complex 3D Printed Surfaces using Filter Response Analysis" has been accepted at the IEEE VR 2021 workshop on on Novel Input Devices and Interaction Techniques (NIDIT). In this work demonstrate the usage sensing touch on arbitrary 3D printed geometries using filter responses in combination with deep neural networks. The paper can be found here.
This research center focuses on the question of how the body regulates inflammation in different organs and, to this end, develop a specific imaging methodology. In our project we will explore multi-scale and multimodal data integration into generative deep neural networks to identify new and explainable machine learning algorithms. For more information please visit the SFB website and for more details about our project can be found here.
CVMLS | | CRC Intelligent Matter is funded by the DFG!
The CRC „Intelligent matter: From responsive to adaptive nanosystems“ is inspired by the question whether synthetic matter can provide artificial building blocks to enable intelligent material properties. In our project we will investigate nanophotonic building blocks and their properties for neural networks. More information can be found here.
CVMLS | | Deep Farm Bots innovation network started!
The recently approved ZIM cooperation network Deep Farm Bots has been started. The central goal of the network is to develop and disseminate new agricultural robotics solutions for efficient and sustainable agriculture. In an interdisciplinary approach, agricultural robotics is to be linked with new methods of Deep Learning and the synergy effects between the partners are to be deepened. The network is funded by the Central Innovation Programme of the Federal Ministry for Economic Affairs and Energy. More information can be found here.
CVMLS | | Paper accepted at the ICPR 2020 Workshop on Explainable AI
Our paper "The Impact of Activation Sparsity on Overfitting in Convolutional Neural Networks" has been accepted at the International Conference on Pattern Recognition (ICPR 2020) workshop on Explainable Deep Learning - AI. In this work we introduce a novel measure for neural activation sparsity along with several visualisation techniques for this property. Using these new insights we demonstrate how sparsity is correlated with overfitting, which intrinsically limits the learning capabilities of the models. The paper can be found here.
CVMLS | | Paper accepted at the VAIB2020 ICPR workshop
Our paper "Towards Visual Insect Camera Traps" has been accepted at the VAIB 2020 workshop which is part of the International Conference on Pattern Recognition (ICPR 2020). In this work we evaluate different deep learning strategies for small object detection in cluttered scenes such as insects in natural environments. The underlying insect in the wild dataset was generated in a collaborative effort with the Institute for Landscape Ecology (University Münster) and the Centre for Biodiversity Monitoring at the Zoological Research Museum Alexander Koenig in Bonn (Germany). Additional information can be found here: paper & video.
Our paper "Embedded Dense Camera Trajectories in Multi-Video Image Mosaics by Geodesic Interpolation-based Reintegration" has been accepted at the WACV 2021 conference. In this work we are presenting a novel algorithm to generate global large-scale image registrations with embedded dense camera trajectories. Using these algorithm we demonstrate that more than 100.000 images can be registered in reasonable time and the resultant reconstructions are widely applicable to land surveys, tracking approaches or large-scale area reconstructions. For more information you can find the publication here.
CVMLS | News article about our research is featured in DIE WIRTSCHAFT MS
Our medical training research project is featured in DIE WIRTSCHAFT Münster & Münsterland. In this project we are investigating the usage of virtual and augmented reality (VR & AR) in combination with advanced user interaction metaphors and machine learning to derive novel interactive teaching strategies. If you are interested in our working group for VR and AR in Biomedical Computer Vision see here and information about the project can also be found here. The article can be found online here.
CVMLS | Teaching Award of the Faculty of Mathematics & Computer Science
We recently won the Teaching Award of the Department of Mathematics and Computer Science at the Universtiy of Münster. Many thanks for this great award. For more information please see here.
Our AI Starter Grant "Learning from Neuroscience to Investigate the "IQ" of Deep Neural Networks" has been accepted! In this project we will investigate strategies from neuroscience to explain deep neural networks. For more details see here.
Our pre-print "Exploiting the Full Capacity of Deep Neural Networks while Avoiding Overfitting by Targeted Sparsity Regularization" is now available on arXiv. In this work we investigate novel explainable AI (XAI) methodologies for deep convolutional neural networks (CNNs). The pre-print can be found here.
Our AI@WWU workshop will start this month. In this course we will teach AI and machine learning basics and apply these techniques to interdisciplinary research problems. More details can be found here.
Pascal Kockwelp joins the group as a PhD student. He will investigate the use auf augmented and virtual reality systems for medical education. Welcome Pascal!
PhD students received a funding for their interdisciplinary pilot project. In this project four scientists from computer science, biology and medicine are developing an imaging system to assess sperm motility - an important factor in determining male fertility. Congratulations to Matthias Kiel, Sebastian Thiele, Samuel Young and Lars Haalck. For more details see here.
Our abstract has been selected to be presented at the Conference DhD 2020 "Spielräume". The talk is entitled: "How deep learning-based images analysis and cultural historical heraldry benefit from each other". Thanks to everybody involved!
Our paper "Towards image-based animal tracking in natural environments using a freely moving camera" has been accepted in Journal of Neuroscience Methods. You can find the article here.
CVMLS | VIRTUAL & AUGMENTED REALITY FOR BIOMEDICAL COMPUTER VISION
Working group Virtual and Augmented Reality for Biomedical Computer Vision: We aim to study and implement innovative VR and AR strategies for novel biomedical computer vision applications. For more details see here.
Our KMU Inno grant "Developing Intelligent 3D Printers (SMARTPRINT)" has been accepted! In this project we will develop novel computer vision and machine learning strategies to improve FDM 3D printing.
Our Start-Up transfer.NRW grant was funded by the European Regional Development Fund and the Ministery of Economy, Innovation, Digitisation and Energy (NRW). This academic transfer funding will support the development of insect tracking systems within the academic spin-out Qubeto GmbH. Congratulations!
Our manuscript and talk "Quantifying the Behavioural Dynamics Behind the Sixth Mass Extinction of Insects" has been accepted at this years EnviroInfo 2019 conference in Kassel. Lars Haalck will present his progress on our insect tracking system. Congratulations to everybody involved! More information can be found here.
CVMLS | PAPER ACCEPTED IN IEEE/ACM TRANS COMP BIO!
Our paper "Resolving colliding larvae by fitting ASM to random walker-based pre-segmentations" has been accepted in IEEE/ACM transactions on computational biology and bioinformatics. You can find the article here.
Our paper "From skylight input to behavioural output: a computational model of the insect polarised light compass" has been accepted in the PLoS computational biology. You can find the article here.
We are very pleased that our insect tracking project is now funded by the Microsoft AI for Earth initiative. In this project we are investigating the limitations of current visual tracking algorithms for very small objects in cluttered environments and aim to develop novel strategies for ecological research.
We are invited to present our work in in-field animal tracking at the BBC in Manchester in April 2019. In particular we will speak about "Detecting and Tracking Animals in Complex Natural Environments".
Our paper "The sulfite oxidase Shopper controls neuronal activity by regulating glutamate homeostasis in Drosophila ensheathing glia" has been accepted in the Nature Communications. You can find the article here.
CVMLS | PAPER ACCEPTED AT MEASURING BEHAVIOR 2018!
Our paper "Possibilities, Constraints and Limitations of Image-based Animal Tracking in Natural Environments" from Benjamin Risse, Michael Mangan and Barbara Webb has been accepted at the Measuring Behavior 2018 (MB2018) conference.
Our paper "Automatic non-invasive heartbeat quantification of Drosophila pupae" from Dimitri Berh, Aaron Scherzinger, Nils Otto, Xiaoyi Jiang, Christian Klämbt and Benjamin Risse has been accepted in the Computers in Biology and Medicine.
