Hosting University: Pázmány Péter Catholic University (Hungary)
Title of seminar: The part and the whole: relations in the scene of sensors’ measurement space
It is my conviction that a system containing a large number of sensors can provide much more formation and knowledge than the simple summation of single sensors. During my earlier career in both industry and academia my favorite activities were around and about intelligent sensor systems. Today, it is clear that the time of smart, self-organized sensor systems has arrived, and I keep pursuing answers and solutions on how to face the related challenges.
Hosting University: Universidad Autónoma de Madrid (Spain)
Title of seminar: Deep Learning for Computer Vision
Deep learning has revolutionized how we solve computer vision problems. The Deep Learning for Computer Vision Workshop that we held for the IPCV scholars over four days in March 2019 introduced students to some of the most important and exciting advances, both theoretical and applied. Students attended 16 lectures in total, covering topics like convolutional networks, recurrent networks, optimization, self-supervised learning, and interpretability. The course was well-received, and the IPCV scholars were bright and enthusiastic.
Hosting University: Universidad Autónoma de Madrid (Spain)
Title of seminar: Research and Innovation in Computer Vision from a technology transfer view-point
I really enjoyed my time with such a multi-cultural group of students. As an entrepreneur with a PhD in Computer Vision, I could offer them two different career paths, and how to eventually merge them. My first session with them was a technical one, focused on photogrammetry and 3D reconstruction techniques, with a focus on industrial applications. The students quickly understood the fundamentals of this field and proposed improvements using newer techniques to improve the results. My second session was about how to start your own business around your research area. Lastly, I had a session with them to solve any questions they would have, and I was gladly surprised that many of them considered starting their own company. In general, it was a great experience and I would totally recommend it.
Hosting University: Université de Bordeaux
Title of seminar: Topics in Machine Learning for Computer Vision: From Inverse Problems to Deep Learning
Being a guest scholar at the Université de Bordeaux has been a very positive experience. I greatly enjoyed my interaction with the students of the IPCV master program. During my visit, I interacted with the students in two formats. The first one was tutoring, in which we discussed the students’ master projects in a free and open environment that encouraged everyone to participate – this included flash talks from the students followed by an interactive discussion. The second form of interaction was through a lecture series; in which I drew attention to the evolution of computer vision from the inverse problem perspective to current developments in deep learning. We navigated through the challenges involved in the numerical realisation of several optimisation models (model based techniques) and how they can be combined with deep nets. We explored several case studies ranging from computational photography for single image reflection removal to recent developments in deep semi-supervised learning for large scale classification problems.
Hosting University: Université de Bordeaux
Title of seminar: Surgical Video Processing and Retrieval
The processing of surgical videos with computer vision methods is a highly relevant topic, which enables to automatically find semantics, such as instruments, surgical actions, diseases, etc. The detection and further analysis of these semantics can help surgeons during the intervention, but also be beneficial for post-operative use cases, where clinicians want to find specific segments and use it for documentation, retrospective analysis, or teaching. In this course, I introduced the IPCV students to this hot topic and discussed several aspects of it. The course was well-received, the students were enthusiastic and contributed great discussions, questions, and interesting comments.
Hosting University: Université de Bordeaux
Title of seminar: Challenges in imaging and sensing in photon-starved regimes
The recent development of highly sensitive sensors operating and extremely high frame/data rates has opened new avenues and opportunities for robotics and computer vision applications. Of primary interest in this Masterclass were single- photon detectors that can be used for Lidar-based 3D imaging in extreme environments and event-cameras that are tailored to analyse fast moving objects. The development of such sensors comes with its own challenges, including high data rates and high uncertainties, that require novel algorithmic structures to be proposed. I introduced the IPCV students to this exciting domain of research topic and discussed how approximate Bayesian methods can be used for scalable uncertainty quantification in such large-scale imaging problems.
The course was well-received. I particularly appreciate the enthusiasm of the students of the IPCV master program and the great offline discussions that followed.