“Signal Processing in the AI era” was the tagline of this year’s IEEE International Conference on Acoustics, Speech and Signal Processing, taking place in Rhodes, Greece.
In this context, Brent de Weerdt, Xiangyu Yang, Boris Joukovsky, Alex Stergiou and Nikos Deligiannis presented ETRO’s research during poster sessions and oral presentations, with novel ways to process and understand graph, video, and audio data. Nikos Deligiannis chaired a session on Graph Deep Learning, attended the IEEE T-IP Editorial Board Meeting, and had the opportunity to meet with collaborators from the VUB-Duke-Ugent-UCL joint lab.
Featured articles:
- De Weerdt, B. et al., “Designing Transformer Networks for Sparse Recovery of Sequential Data Using Deep Unfolding” (https://ieeexplore.ieee.org/abstract/document/10094712)
- Yang X. et al., “Relevance Propagation through Deep Conditional Random Fields” (https://ieeexplore.ieee.org/abstract/document/10095075)
- Stergiou A. et al., “Play it back: Iterative attention for audio recognition” (https://ieeexplore.ieee.org/abstract/document/10096532, developed at the University of Bristol)
On August 31 2021 at 13.00 Gobinath Jegannathan will defend his PhD entitled “Current-assisted SPAD sensors fabricated in conventional CMOS”.
Everybody is invited to attend the online presentation via https://us02web.zoom.us/j/88093194180?pwd=ODkxazlYazRGeXFCSDNqRStSY2VZZz09 (Meeting ID: 880 9319 4180 Passcode: 35fKsq)
A major revolution in light-detection is on-going. Whereas in the past, the electrical signal from light incident on a detector had to be amplified with a permanent dissipative analog amplifier, today there exist single photon avalanche diodes (SPADs) that natively detect single photons. Detecting the smallest measure of light is the ultimate achievable feat of photodetection. Although one might think that single-photon detectors are niche scientific instruments for fundamental research, it is rather a tool used in many day-to-day applications. 3-D image sensors, PET scanners and proximity sensors in mobile phones all use single-photon detection.
Moreover, SPADs are single-photon detectors which are very popular because they can be mass-fabricated in CMOS technology. The recent demonstration of a 1-megapixel SPAD from EPFL and Canon testifies the advancement of SPAD sensors, giving insight in the imaging possibilities yet to come.
In this work, a novel SPAD sensor is presented where the novelty arises from the integration of a large absorption volume and a very small avalanche multiplication volume. Such a detector topology allows to have a thick absorption layer which leads to higher quantum efficiencies for NIR wavelengths. This integration is enabled by “current-assistance” principle where a drift field is created in the substrate by applying a potential gradient. This “current-assisted SPAD” is fabricated in a cost-effective CMOS process which is commercially available.
On August 29 2022 at 14.00, Lusine Abrahamyan will defend her PhD entitled “Optimizing Deep Learning Methods for Computer Vision”.
Everybody is invited to attend the presentation live (in room D0.0.8) or online via this link.
In the past decade, the interest in intelligent applications, ranging from smart homes and healthcare to social networks and autonomous driving, has drastically increased. This has led to significant progress in machine learning research. Notably, deep learning, a subfield of machine learning, has gained popularity due to its superior performance in numerous computer vision or natural language processing tasks. For instance, deep learning based models trained on large-scale datasets can determine infrequent and interesting collision events in the dataset collected using a Large Hadron Collider at CERN. Despite this advancement, there are still challenges that need to be addressed to fully harness the potential of deep learning methods. This thesis focuses on three such challenges: democratizing distributed learning, tackling task-specific problems during the model optimization process and designing deep learning architectures for mobile devices. The challenge of performing distributed learning is the cost of transferring a huge amount of information at each iteration of the training. This problem becomes worse when distributed learning is performed through a wireless network due to limited bandwidth. The next challenge concerns the process of model optimization. There can be specific problems in every task that need to be handled. For example, if in a classification dataset, the number of images of one class is significantly higher than that of another class, the model would generalize poorly due to the class-imbalance problem. Thirdly, as mobile devices are an integral part of our lives, the design of high-performance deep learning models for such devices is crucial. This includes a design of architectural modules that can efficiently utilize the learnable parameters in order to provide the highest possible increase in performance. Taking a step in addressing these challenges, our first contribution in this thesis is a novel framework for distributed training. Our solutions employ a lightweight neural network to compress the information that is being sent at every iteration to reduce the communication bandwidth. The proposed framework can reduce the amount of transmitted information up to 877x. Our second contribution is the introduction of two loss functions designed to tackle problems arising in image classification and single image super-resolution tasks. Finally, our third contribution is the development of a new family of compact models for on-device inference and the efficient architectural unit for the task of real-time semantic segmentation.
After two years of dedicated research and development under the leadership of ETRO-VUB, a breakthrough has been achieved within the INTOWALL project: a revolutionary radar technology for building inspection was developed, called the transient radar method (TRM). The initiative aimed to reduce the CO2 emissions of buildings and increase their energy efficiency.
The new technology enables the measurement of the density of glass wool in cavity walls with unprecedented precision, without the need for invasive methods. “This advancement not only promises to improve the accuracy of insulation assessments but also contributes to the ambition to achieve a CO2-neutral status by 2050,” says Professor Johan Stiens of ETRO.
Looking towards the future, the project team is focused on further refining the technology to map a wide range of insulation materials and building elements. This prospect of expansion and application on a larger scale highlights the endless possibilities. As part of the FTI Brussels Festival, the milestone of the INTOWALL project will be celebrated. A unique demonstration was held on March 18, 2024.
Additionally, the project team invites potential partners to contribute to and participate in this groundbreaking endeavour. Through collaboration, we can transform the construction sector into a more sustainable and efficient future.
For more information on InToWall press articles: https://press.vub.ac.be/wereldprimeur-in-radartechnologie and https://trends.knack.be/kanaal-z/z-nieuws/bekijk-radar-van-vub-ziet-isolatie-dwars-door-muren-heen/
On Monday September 19, Prof. Nikos Deligiannis, Prof. Bruno Da Silva and Prof. Bart Jansen gave a warm welcome to the new generation of MACS students. We expect 50+ new students in the first master year.
His speach is entitled “Klimaatverandering in een historisch, mondiaal en antropogeen perspectief“, or “climate change in a historical, mondial and antropogene perspective”. In this short introduction, we look at the current climate change at a larger time scale and across the different regions of the world. We highlight some of the relationships with the world population increase and the local and mondial economic activity, focusing on energy consumption. We finish the talk with a vision on the future in the domains of energy sources and transitions in automotive
More info at:
Looking for ways to scale your current and future IT needs? Thoughts about Cloud migration and how to best balance pros and cons? Eager to learn how open-source technology fits in?
Join the TETRA OpenCloudEdge seminar, which will take place Thursday afternoon, 24 February, in a hybrid physical & online format.
Industry and academic speakers will present experience with OpenStack, Kubernetes and related open-source cloud technologies. They will discuss advantages and pitfalls as well as presenting opportunities.
The physical venue location is DSP Valley, Esperantolaan 4, 3001 Leuven. A link for online participation will be sent to registrants.
Find out more about the program and register: https://share.hsforms.com/1YhpS0Nt_RpmUxV1Zzh2XAw42q9f
