“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 April 25th 2025 at 16:00, Ayman Morsy will defend their PhD entitled “A NOVEL APPROACH TO DEPTH-SENSE IMAGING USING CORRELATION-ASSISTED DIRECT TIME-OF-FLIGHT”.

Everybody is invited to attend the presentation in room I.0.03 or online via this link.

Time-of-flight (ToF) imaging has emerged as a vital technology in machine vision and sensing, expanding into applications such as augmented and virtual reality, gaming, robotics, autonomous driving, autofocus, and facial recognition on smartphones and laptops. ToF technology determines the distance to an object within the detection range by emitting a light source and measuring the time it takes to return. This round-trip time determines the object’s distance, with different sensing technologies employing distinct methods to determine this time.

For ToF applications, developing sensors with high image resolution, low power consumption, and the ability to function reliably in high ambient light conditions is desirable. This dissertation presents the development of a novel single-photon avalanche diode (SPAD)-based pixel called Correlation-Assisted Direct Time-of-Flight (CA-dToF), designed for in-pixel ambient light suppression and characterized by low power consumption and a scalable pixel structure. The CA-dToF pixel uses a laser pulse correlated with two orthogonal sinusoidal signals as input to two switched capacitor channels, which average out detected ambient light while accumulating the laser pulse round-trip time.

To gain insights into CA-dToF pixel operation, both Python simulation and analytical modeling were developed. Two generations of the CA-dToF pixel were developed and characterized, with the second-generation pixel achieving the first operational performance under high ambient light conditions. The two-generation CA-dToF pixel was tested under various lighting conditions and pixel design variations. Additionally, noise sources within the pixel implementation were analyzed, and potential solutions were proposed.

Dear [Newcomer],

Welcome to the ETRO.RDI LARGE research group! We are excited to have you on board.

Your colleagues will be happy to assist you as you settle in. However, they may not have all the answers—that’s where we, the central support team of ETRO.RDI, come in.

At ETRO.RDI, we believe that getting to know each other makes everything smoother. Collaboration becomes more efficient, and let’s be honest—it’s just more fun!

We understand that integrating into a new team takes some effort, but the sooner you start, the quicker you’ll feel at home.

Let’s avoid a bumpy start; please complete the following steps smoothly.

Immediately

✅ Introduce yourself – Share three short bullet points with personal details you’d like to share with the ETRO team. These will be used for your introduction via ETRO.TV, one of our internal communication platforms.

Once You Receive Your ETRO Credentials

✅ Log in to the intranet – Visit www.etrovub.be and click the icon in the top-right corner to log in. Take some time to explore the platform.

✅ Complete your personal ETRO.RDI webpage – Log in at https://www.etrovub.be/wp-admin/ and update your profile using your etrovub username and password (without adding “etrovub\”).

✅ Schedule your web picture appointment – Once you are in the country, contact Luc van Kempen to arrange your professional photo session. This ensures a uniform presentation of yourself and your research topic on our website and internal communication channels.

By completing these steps, you’ll make it easier to connect with colleagues. The first time you meet someone, chances are you’ll already know a little about each other—perhaps even sharing a common interest you’ve mentioned earlier.

---

If you ever feel unsure or need assistance, don’t hesitate to reach out (see the contact details below).
We’re here to support you! Sometimes, we’ll help you directly, and other times, we may refer you to the intranet or connect you with a colleague who has encountered a similar challenge.

📧 For research-related or administrative questions:

Karin De Bruyn, ETRO Research Coordinator – Karin.Debruyn@vub.be

💻 For IT-related support:
ICT@etrovub.be – We’ll assist you directly or forward your request to the central helpdesk if needed.

🔧 For technical, purchase or access issues:

Luc van Kempen, ETRO System Manager –  Luc.van.Kempen@vub.be We look forward to meeting you!

Dear Future User of ETROFARM,

To process your access request, please provide the following information:

• Your Last Name
• Your First Name
• A valid email address for communication (preferably your VUB email; using it for the request is ideal)
• Your Principal Investigator’s (PI) name (adding them in CC is ideal)
• Your public SSH key in OpenSSH format (2), which will be used to connect to ETROFARM

Steps to follow:

1. Send your access request to ICT@etrovub.be
2. You will receive an initial email containing your ETROVUB credentials, allowing you to:
3. Install OpenVPN (1) (if not already installed) and connect to the ETRO Bubble
4. Receive a second email confirming your access to ETROFARM
5. Connect to ETROFARM (3)
6. Read the provided documentation (4)

This is our standard procedure. If you’ve already provided some of the requested information, there’s no need to resend it.

Kind regards,
ict@etrovub.be

1.    Install and use OpenVPN

To access machines in the ETROVUB bubble (using an etrovub account), please use an OpenVPN connection.

In order to download the software, surf to https://vpn.etrovub.be and login with your etrovub credentials.

Download the latest software it offers to you and install.

Once done you can open it and connect with the same etrovub credentials (username only, no etrovub before or after), and click ok.

The connection establishes and shows:

2.    Generate RSA 2048-bit OpenSSH keys

When generating SSH keys, the private key should always be stored securely on your local machine,

while the public key is meant to be shared—simply copy it and send it to the other party.

You’ll also be asked whether to use a passphrase: this is optional.

It’s recommended to use a passphrase on laptops or shared systems for extra protection,

but you can skip it if you’re setting up automation or scripts where prompts would get in the way .

Here’s how to generate an RSA 2048 OpenSSH key pair on Windows, macOS, or Linux:

Windows (with PuTTYgen)        

• Open PuTTYgen
• Select:

• Type: RSA
• Bits: 2048

• Click Generate → move mouse
• (Optional) Set a Key passphrase
• Copy public key from ‘Public key for pasting into OpenSSH…’

• send to the other party
• Click Save private key → store as id_rsa.ppk locally

MacOS (Terminal)

ssh-keygen -t rsa -b 2048 -f ~/.ssh/id_rsa

• Hit Enter to skip passphrase for automation
• Show public key:
• cat ~/.ssh/id_rsa.pub → send to the other party
• Private key stored: ~/.ssh/id_rsa

Linux (Terminal)

Same as macOS:

ssh-keygen -t rsa -b 2048 -f ~/.ssh/id_rsa

• Skip passphrase if automating
• Show public key:
• cat ~/.ssh/id_rsa.pub → send to the other party
• Private key stored: ~/.ssh/id_rsa

3.    Connect to ETROFARM

login using ssh or equivalent

on: etroflock.etrovub.be (10.0.5.202)

using your

• credential login (without the @…)
• private key.

4.    Why the GPU Farm at ETRO?

At ETRO, advancing cutting-edge research hinges on the effective use of High-Performance Computing (HPC) resources. We recognize the need to bridge the gap between desktop computing limitations and the extensive power of HPC systems like Hydra. Our new dedicated GPU farm is designed to transform how users engage with computing resources—moving from power-hungry, noisy desktops in overheated rooms to a more flexible, efficient, and scalable solution.

This upgrade is not just about new hardware; it’s about reshaping user behavior and enhancing integration with VUB’s existing HPC platforms. By aligning the GPU farm with VUB’s systems, we ensure smoother project transitions and better compatibility with Hydra, allowing researchers to focus more on innovation rather than infrastructure.

The GPU farm will help ETRO achieve cost-efficiency, reduce our environmental footprint, and support research with greater flexibility and computational power—all without the high costs associated with commercial cloud solutions or the limitations of local desktops.

Available Resources

Hardware

We offer three nodes with the following configurations:

Partition	Node	CPU	GPU
FARM	ETROFARM	AMD EPYC 9124 (F. 25 M.17) 2 sockets 2x16x1= 32 (logical) CPU 377GB	4x Nvidia A100 (Ampere)   Nvidia Driver: 535.183.01   CUDA: 12.2   80GB
COOP	ETROCOOP01 ETROCOOP02	INTEL 13th Gen Intel(R) Core(TM) i9-13900 (F.6 M.183)   1 socket 1x8x2= 16 (logical) CPU   125GB	2x Nvidia GeForce RTX4090 (Ada Lovelace)   Nvidia Driver: 535.183.01   CUDA: 12.2 24GB

Additional Information:

• Data is accessible via OPENVPN.
• Network transfer rate: 30MB/s.
• Diskspace: 1TB workspace provided, with the possibility of creating a group workspace upon request. Diskspace is shared over a 1 Gbps network.

Software

We provide software built with Easybuild and managed with Lmod. While we strive to offer packages similar to those available on Hydra, our node architecture and resources are different. We aim to include the most common packages and provide configurations similar to those found at Hydra via Easybuild

Easyconfigs

Users can supply their own build files if necessary.
Note: Singularity containers (as used in Hydra) are not provided.

Priority

We use SLURM’s job accounting and fairshare system to manage resource allocation and prevent monopolization.
The fairshare score reflects cluster usage and helps prioritize jobs.

Training Resources

For SLURM functionality, see the theoretical guide at SLURM:
https://slurm.schedmd.com/.

Practical documentation closely aligned with our setup can be found at Hydra Documentation:
https://hpc.vub.be/docs/

Hydra also offers regular training sessions.

Two of ETRO’s postdocs, Angel for the project Equitable Oxymetry and Abel for the I-Healthy path project, have been selected for the MedTech accelerator: https://lifetech.brussels/en/medtech-accelerator-en/.

More info and pictures in these links:

• https://www.linkedin.com/posts/lifetech-brussels_medtechaccelerator2025-value-patients-activity-7305982232211476481-9Sh9?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAApXK0BavQNkWiiuuorRLD6_ShNleRksEs
• https://lifetech.brussels/en/medtech-accelerator-en/

Brussels, March 8, 2025 – On International Women’s Day, a day dedicated to celebrating women’s achievements and advocating for greater gender equality, the new WATS ambassador is announced: Lesley De Cruz. This inspiring professor at the Vrije Universiteit Brussel (VUB) and researcher at the Royal Meteorological Institute of Belgium (RMI) is committed to encouraging girls and women to pursue careers in STEM (science, technology, engineering, and mathematics). 

The WATS Award: Increasing Visibility for Women in STEM 

Since 2018, the Brussels-Capital Region and Innoviris have been presenting the biennial Women Award in Technology and Science (WATS). This award highlights Brussels-based women who not only excel in STEM but also inspire others to follow in their footsteps. The prize is awarded based on their potential as role models and their concrete plans to encourage girls and women to consider STEM studies or careers. 

Why Lesley De Cruz? 

Lesley De Cruz was selected by a jury of former laureates and young female STEM enthusiasts (aged 16-17) for her versatility as a scientist and her approachability as a role model. She also excels in her ability to communicate complex scientific topics in an accessible and engaging way. 

Her award-winning project introduces young people to artificial intelligence by guiding them to build a virtual city and examine its climate impact. This innovative approach makes scientific knowledge both relevant and tangible while addressing societal issues such as climate change. 

Making STEM Accessible for Everyone 

Lesley’s mission as a WATS ambassador is clear: “I want to show girls that STEM empowers you to make a positive impact on the world,” she says. “My message is especially aimed at girls who think, ‘Amazing, but that’s not for me.’ I want to inspire them to believe in themselves and show that anyone, regardless of background, can have a future in STEM.” 

With the accompanying €10,000 prize, Lesley plans to organize various activities in Brussels to actively guide and motivate girls. 

International Women’s Day: Celebrating Female Role Models 

On this International Women’s Day, we celebrate not only the achievements of women worldwide but also the role models who make a difference. Thanks to Lesley De Cruz and her efforts, we are building a future where STEM is a place for everyone, regardless of gender. 

On March 24th 2025 at 16:00, Selene De Sutter will defend their PhD entitled “SEGMENTATION OF GLIOBLASTOMA FROM MULTI-MODAL MEDICAL IMAGING – Towards revealing tumor infiltration”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Glioblastoma is the most aggressive type of primary brain tumor, and its treatment relies on accurately identifying the tumor’s boundaries – a process known as segmentation. However, manually outlining these areas is labor-intensive and can be susceptible to variability between experts.

Tumor boundaries are typically defined using magnetic resonance imaging (MRI), but glioblastoma cells often spread beyond what is visible on these scans, further complicating the segmentation process. These hidden, infiltrating cells are difficult to detect, often going untreated and contributing to tumor relapse. Detecting them is therefore crucial for improving treatment outcomes. Positron emission tomography (PET) with amino acid tracers may provide additional information about these infiltrating cells, but automated segmentation on these images remains underexplored.

This work investigates how image analysis and artificial intelligence techniques can enable accurate and automated segmentation of glioblastoma from medical images. The proposed contributions are initially focused on segmentation of the known tumor boundaries on MRI, while examining how different types of MRI scans contribute to this task. The focus then shifts towards segmentation using PET imaging, where we highlight its current limitations. Finally, we combine information from both MRI and PET scans to achieve a more comprehensive and accurate tumor segmentation, which may contribute to improved treatment planning and outcomes.

On February 20th 2025 at 16:00, Ángel Solé Morillo will defend their PhD entitled “M3-PPG: TOWARDS NOVEL (PERSONALIZED) PHOTOPLETHYSMOGRAPHY SYSTEMS THROUGH THE UNDERSTANING OF KEY INFLUENCING FACTORS”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Photoplethysmography (PPG) is a low-cost technique that allows for extracting physiological parameters, such as heart rate or blood oxygen saturation, through light interactions with the skin. PPG has been present in clinical practice as the technology behind pulse oximeters since the 1980s. With the proliferation of health wearables equipped with PPG sensors in the last 15 years and the advancement in PPG applications beyond pulse oximetry, a new perspective has arisen, with PPG having the potential to tackle some key societal and health issues of the 21st century.

Despite its widespread adoption, PPG remains susceptible to various factors that can compromise the accuracy of the physiological measurements. Understanding these influences individually can improve and expand the use and applications of PPG, ultimately enabling personalized health monitoring.

This research first proposes a theoretical framework, which describes key hardware and software improvements that can enable robust and personalized physiological monitoring using PPG technology. Next, an analysis of the impact of the instrumentation on the PPG signal is presented.

In addition, the impact of skin tone on the PPG signal is evaluated. Melanin, another PPG influencing factor whose content in the skin gives rise to different skin colors, is measured with two prototypes in a pilot study. This contributes to addressing the oxygen saturation overestimation in pulse oximeters for users with darker skin tones.

The final part integrates all previous research findings into a prototype designed for continuous vital sign monitoring at the chest, whose performance was validated through an initial pilot study with healthy participants. This work also analyses how regulations impact this prototype’s possible road to market as a medical device in Europe and the U.S.

On February 5th 2025 at 16:00, Raees Kizhakkumkara Muhamad will defend their PhD entitled “COMPRESSION STRATEGIES FOR DIGITAL HOLOGRAPHY”.

Everybody is invited to attend the presentation in room I.0.01 or online via this link.

Holographic techniques sample a 2D interference pattern produced by coherent light waves reflected or transmitted from the different objects in the 3D scene. As a display technology, it provides all necessary visual cues for perceiving the scene by the brain without causing mismatches between accommodation and vergence of the eyes. Non destructive imaging with high resolution for biomedical and industrial inspection also utilizes holographic principles. Holographic microscopes are realizable with optically simpler setups than regular microscopy, opening new pathways for computational microscopy. Utilizing more complex arrangements, such as holographic tomography, allows for reconstructing the 3D refractive index profile of transmissive objects, resolving even sub-cellular structures with visible light. It represents the culmination of humanity’s effort to record and represent light information.

However, sampling interference patterns for high-end displays or highresolution microscopy result in an extensive digital footprint. Historically, for many multimedia use cases, the data transmission bottleneck dictated the fidelity of the consumed content, and one can expect holography to be no different. Compression algorithms can help mitigate the data load, trading off more computation for an effective increase in transmission capacity. The algorithms must be tailored for holograms used in practice and exhibit a computational complexity appropriate for the use case, particularly on the decoder side. This thesis presents compression strategies for effectively tackling such use cases for holography.

The performance of conventional image compression tools on metrological hologram content is first studied. We provide a novel analysis of compression artifacts on the retrieved metrological data obtained for digital holographic microscopy and tomography. First-generation holographic displays are poised to use binary representation due to difficulties in modulating pixels at the sizes required by holography. Context-based compression is adequate for lossless and near-lossless compression of such data. Here, one extracts from a pixel the redundant information from previously decoded neighbouring pixels using a generalized Markovian model. Another context-based framework utilizing linear (autoregressive) models is used to design a highly scalable lossless compression scheme for non-binary holographic data. For broadcast scenarios, it is desirable to have a compression system that can support view-selective decoding to minimize the transmission of unutilized information. For this purpose, we propose a short-time Fourier transform (STFT) based codec, which slices the hologram into independently decoded, spatio-angular chunks. Given a target mean-squared error, the optimization techniques efficiently round down the signal in the STFT domain by application of adaptive quantizers. All these compression schemes are applicable for a single frame of hologram data and feature lightweight decoding architectures while surpassing compression performances achieved by any existing solution on most tested holograms. To compress holographic videos with arbitrary motion, we utilize a novel motion compensation algorithm that can predict rotational motion in conjunction with the above-mentioned STFT framework. The Markovian and STFT frameworks discussed in this work have been adopted as part of the first international hologram compression standard, JPEG Pleno — Part 5: Holography (ISO/IEC 21794-5)

ETRO, the Vrije Universiteit Brussel (VUB) and imec are proud to announce that Prof. Nikos Deligiannis has been awarded a prestigious ERC Consolidator Grant from the European Research Council to make a groundbreaking contribution to science and society.

Project: IONIAN: Reinventing Multiterminal Coding for Intelligent Machines
Budget: €1,999,404

Professor Deligiannis’ IONIAN project focuses on reinventing multiterminal coding, a crucial technology for efficient communication and collaboration between intelligent machines. With the explosive growth of data, such as video and point cloud streams, current storage and communication technologies are under pressure, undermining the ability of intelligent machines to cooperatively perceive their environment. This project develops a groundbreaking compression and communication approach based on interpretable and explainable AI that breaks the limits of traditional compression and cooperative perception techniques.

IONIAN combines classical theories, such as distributed source coding, with modern deep learning techniques and explainable AI, focusing on three innovative pillars:

1. Solving theoretical limitations in multiterminal coding.
2. Designing interpretable and efficient machine learning models.
3. Developing new methods to enhance model interpretation.

The goal of this project is to elevate the collaboration between intelligent systems, such as autonomous vehicles and mobile robots, to a higher level, with greater safety and trust as the result.

Remote editing with VS Code

Visual Studio Code is a widely-used, cross-platform Integrated Development Environment (IDE) that supports numerous programming languages and offers a vast array of extensions to enhance its functionality.

One notable extension enables development on remote machines via SSH, providing integrated access to a file explorer, terminal, and text editor on the configured remote system. This makes it a strong alternative to JupyterLab as a remote editor for e.g., the ETROFARM Slurm cluster.

Configuring Visual Studio Code extensions for working on ETROFARM Slurm cluster

Open the Extensions tab using the corresponding icon in the left toolbar.

Search for “SSH” and subsequently select the topmost “Remote – SSH” extension. Install this extension.

Depending on your programming language of choice, you might also be interested in extensions such as “Python”, “Python Debugger”, “Ruff” (a Python code linter) etc.

With the “Remote – SSH” extension installed, a new tab “Remote explorer” has been added to the left toolbar.

Add a new remote by pressing the + icon.

When asked for the SSH Connection Command:

ssh <username>@etroflock.etrovub.be

Secondly, it will ask where to store this information. This can be the default option.

The etroflock.etrovub.be remote has been created. Time to connect by triggering one of the 2 corresponding buttons.
During a brief instant the option will be displayed to edit the configuration. If you have missed it you can find this file via the gear button next to remote explorer – remote tunnels – ssh

Your config file should look like

Host etroflock.etrovub.be
  HostName etroflock.etrovub.be
  User jdoe
  IdentityFile C:\users\jdoe\.ssh\id_rsa

If you have no experience with encryption you can e.g. use a rsa 2048 type of key. Please make sure you are using a private key in openssh format.

Upon our first connection attempt, it requests the platform of the remote host, being the Slurm cluster’s login node we are connecting to. This is a Linux machine.

It will also ask to confirm the SSH public key credential of the server.

We are now connected to the remote server. This can be seen in the Remote explorer tab as well as in the left corner of the bottom toolbar.

Configuring the file explorer and text editor on the remote system

Time to open our file explorer via the “Explorer” tab in the left toolbar (Ctrl + Shift + E). Press the “Open Folder” button. It should by default suggest to open your home folder on the Slurm cluster (currently on /FARM/<username>).

If prompted, confirm that the remote server is (again) a Linux platform. Lastly, confirm that you trust the authors of the files in this folder as this is your own home folder.

Congratulations! Your remote file explorer and text editor on the Slurm cluster is now operational.

Remote terminal and running code on the remote system

A remote terminal can be opened using Terminal -> New Terminal in the top toolbar, or via the Ctrl + Shift + ` shortcut.

An interactive terminal session is opened on the remote host as if it was a PuTTY (or other) SSH session.

Running code remotely with VScode (ideally on Slurm compute nodes)

With the remote file explorer, remote text editor and remote terminal sessions available, it is a logical next step to focus on running our code remotely on the machine. Luckily, this is typically as straightforward as pressing the “Run file” button on the active (Python) file.

We can observe that the code has indeed been executed on the remote machine. However, the configured remote machine is ETROflock, ETRO’s Slurm cluster’s login server that is scarce in compute resources and lacks and GPU’s.

Running our code on a Slurm compute node is more complicated as it involves requesting a Slurm job. This is currently a manual process but we are investigating if this can be automated in VScode by using a custom launch script.

For now, there are two possibilities to run code on the Slurm cluster from within VScode:

A Slurm job can be requested that immediately runs the code until completion (or an error or timeout). This is the recommended default approach for running Slurm jobs.

E.g. the same test.py code is run as a Slurm job by using the srun command.

We observe that the first command is run on ETROFLOCK (the Slurm login node) because that is immediately launched on the remote system. The second command is scheduled as a Slurm job and is run on ETROFARM (a Slurm compute node).

The second possibility involves starting a Slurm job with an interactive shell. Once this interactive shell is running on a compute node, we can manually launch the desired code within that shell. This solution might be preferred when developing and testing the functionality of the code as a Slurm job (and potential queue) must only be requested once per session.

Launching a Slurm job with interactive shell is possible using the following srun command parameters:

srun –pty bash -i

E.g. the same test.py code is run after an interactive shell has been requested using the srun command and has started.

Observe that after requesting the slurm job with interactive terminal we change from “steffen@ETROFLOCK” to “steffen@ETROFARM”. This indicates our interactive shell is indeed running on a compute node instead of the login node.

Executing the test.py script from within this shell again confirms that the code is indeed run on the compute node with hostname ETROFARM.

When finished, one should use the “exit” command. This closes the interactive shell and terminates the Slurm job, thereby releasing the allocated resources for new jobs.

Filezilla is a free and open-source file transfer tool that can be used to exchange files with e.g., ETROFARM Slurm cluster. It requires some brief configuration in order to be able to connect using the SSH public/private key credentials. The required configuration steps are provided in the following tutorial.

Open the Site Manager, via the leftmost icon in the toolbar. Contrary to using the Quickconnect function which only supports connecting via a password, the Site Manager also allows configuring SFTP authentication via SSH keys.

Add a New site with the following configuration:

1. Protocol: SFTP – SSH File Transfer Protocol
2. Port: Optional field, can remain empty (default SFTP value: 22)
3. Host: etroflock.etrovub.be  (IP address: 10.0.5.202)
4. Logon Type: Key file
5. User: Username of your slurm account, typically your ETRO username
6. Key file: Path to your private SSH key used for Slurm
   Note: When browsing, you might need to look for “All files” as there might be no file extension. For UNIX-style systems (Linux / MAC) this is typically found in the user’s home / .ssh directory which is a directory that might be hidden.

Connect to this SFTP server and trust the etroflock’s public SSH key.

All future FileZilla SFTP connections to ETROflock can be easily launched from the Site manager tab.

On November 15th 2024 at 10:00, Eden Teshome Hunde will defend their PhD entitled “CROSS-LAYER DESIGN, IMPLEMENTATION AND EVALUATION OF IPV6 MULTICAST FOR RADIO DUTY CYCLED WIRELESS SENSOR AND ACTUATOR NETWORKS”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Wireless sensor and actuator networks (WSANs), consist of small-sized constrained sensing and/or actuating devices which can run an Internet Protocol version 6 (IPv6) network stack. In such WSANs, there is often the need for sending the same message to several devices in the WSAN, such as for applications for controlling all the lights in a building or turning on/off all the air conditioners in the room etc. For doing so, IP multicast can be an efficient solution.

In this work, we study Bidirectional Multicast RPL Forwarding (BMRF) as this protocol relies on forwarding tables put in place by the well-known Routing Protocol for Low Power and Lossy Networks (RPL) and allows to combine the best ideas of existing multicast protocols. Through RPL, a routing tree towards the sink is installed for multihop routing from node to sink, and the nodes’ forwarding tables will also contain entries for reaching destinations in downward direction.

For downward forwarding IPv6 multicast packets, two methods exist. One is via link layer (LL), broadcasting a frame containing the IPv6 multicast packet. The other is to send several LL unicast frames containing that packet. BMRF allows a node to choose between these two methods. The best option will depend on the presence of a radio duty cycling (RDC) protocol. RDC is part of the medium access control (MAC) layer and puts the radio to sleep when no communication is needed. We investigate the influence of MAC/RDC protocols on BMRF’s performance.

We evaluate the performance of BMRF on non-synchronized WSANs that use Carrier Sense Multiple Access (CSMA) as MAC and ContikiMAC as RDC. We demonstrate that LL unicast outperforms LL broadcast in terms of packet delivery ratio (PDR), delay, and energy consumption in many settings.

We investigate the performance of BMRF on WSANs with synchronous MAC and RDC based on Time Slotted Channel hopping (TSCH). This is more challenging, as TSCH needs a schedule to tell which action must happen in each timeslot. The actions can be to send or to listen on a given channel or to be idle. Idleness allows the radio to switch OFF, providing RDC. The schedule is not part of the standard and must be proposed by the system designer. An elegant autonomous scheduling method called Orchestra is available to accommodate traffic in a RPL tree. We extend Orchestra with a novel scheduling rule for supporting LL downwards forwarding through LL broadcast. Comparing LL unicast with LL broadcast forwarding teaches us that LL unicast outperforms LL broadcast in terms of packet delivery ratio (PDR), but the latter can be beneficial to certain applications, especially those sensitive to delay.

Before conducting the two previous evaluation studies, we investigate the performance of simple convergecast traffic while considering ContikiMAC and TSCH with Orchestra under RPL on the real dual Zolertia Firefly Motes (one is observed and other one is observing mote). This study served two purposes; it reminds the reader of the characteristics of those protocols and allowed to fine-tune the dual motes.

We also contributed by adapting the Orchestra to bursty convergecast traffic. Simulation results demonstrate that the new scheduler slightly improves PDR and reduces delay compared to state-of-the-art solutions.

“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 April 25th 2025 at 16:00, Ayman Morsy will defend their PhD entitled “A NOVEL APPROACH TO DEPTH-SENSE IMAGING USING CORRELATION-ASSISTED DIRECT TIME-OF-FLIGHT”.

Everybody is invited to attend the presentation in room I.0.03 or online via this link.

Time-of-flight (ToF) imaging has emerged as a vital technology in machine vision and sensing, expanding into applications such as augmented and virtual reality, gaming, robotics, autonomous driving, autofocus, and facial recognition on smartphones and laptops. ToF technology determines the distance to an object within the detection range by emitting a light source and measuring the time it takes to return. This round-trip time determines the object’s distance, with different sensing technologies employing distinct methods to determine this time.

For ToF applications, developing sensors with high image resolution, low power consumption, and the ability to function reliably in high ambient light conditions is desirable. This dissertation presents the development of a novel single-photon avalanche diode (SPAD)-based pixel called Correlation-Assisted Direct Time-of-Flight (CA-dToF), designed for in-pixel ambient light suppression and characterized by low power consumption and a scalable pixel structure. The CA-dToF pixel uses a laser pulse correlated with two orthogonal sinusoidal signals as input to two switched capacitor channels, which average out detected ambient light while accumulating the laser pulse round-trip time.

To gain insights into CA-dToF pixel operation, both Python simulation and analytical modeling were developed. Two generations of the CA-dToF pixel were developed and characterized, with the second-generation pixel achieving the first operational performance under high ambient light conditions. The two-generation CA-dToF pixel was tested under various lighting conditions and pixel design variations. Additionally, noise sources within the pixel implementation were analyzed, and potential solutions were proposed.

Dear [Newcomer],

Welcome to the ETRO.RDI LARGE research group! We are excited to have you on board.

Your colleagues will be happy to assist you as you settle in. However, they may not have all the answers—that’s where we, the central support team of ETRO.RDI, come in.

At ETRO.RDI, we believe that getting to know each other makes everything smoother. Collaboration becomes more efficient, and let’s be honest—it’s just more fun!

We understand that integrating into a new team takes some effort, but the sooner you start, the quicker you’ll feel at home.

Let’s avoid a bumpy start; please complete the following steps smoothly.

Immediately

✅ Introduce yourself – Share three short bullet points with personal details you’d like to share with the ETRO team. These will be used for your introduction via ETRO.TV, one of our internal communication platforms.

Once You Receive Your ETRO Credentials

✅ Log in to the intranet – Visit www.etrovub.be and click the icon in the top-right corner to log in. Take some time to explore the platform.

✅ Complete your personal ETRO.RDI webpage – Log in at https://www.etrovub.be/wp-admin/ and update your profile using your etrovub username and password (without adding “etrovub\”).

✅ Schedule your web picture appointment – Once you are in the country, contact Luc van Kempen to arrange your professional photo session. This ensures a uniform presentation of yourself and your research topic on our website and internal communication channels.

By completing these steps, you’ll make it easier to connect with colleagues. The first time you meet someone, chances are you’ll already know a little about each other—perhaps even sharing a common interest you’ve mentioned earlier.

---

If you ever feel unsure or need assistance, don’t hesitate to reach out (see the contact details below).
We’re here to support you! Sometimes, we’ll help you directly, and other times, we may refer you to the intranet or connect you with a colleague who has encountered a similar challenge.

📧 For research-related or administrative questions:

Karin De Bruyn, ETRO Research Coordinator – Karin.Debruyn@vub.be

💻 For IT-related support:
ICT@etrovub.be – We’ll assist you directly or forward your request to the central helpdesk if needed.

🔧 For technical, purchase or access issues:

Luc van Kempen, ETRO System Manager –  Luc.van.Kempen@vub.be We look forward to meeting you!

Dear Future User of ETROFARM,

To process your access request, please provide the following information:

• Your Last Name
• Your First Name
• A valid email address for communication (preferably your VUB email; using it for the request is ideal)
• Your Principal Investigator’s (PI) name (adding them in CC is ideal)
• Your public SSH key in OpenSSH format (2), which will be used to connect to ETROFARM

Steps to follow:

1. Send your access request to ICT@etrovub.be
2. You will receive an initial email containing your ETROVUB credentials, allowing you to:
3. Install OpenVPN (1) (if not already installed) and connect to the ETRO Bubble
4. Receive a second email confirming your access to ETROFARM
5. Connect to ETROFARM (3)
6. Read the provided documentation (4)

This is our standard procedure. If you’ve already provided some of the requested information, there’s no need to resend it.

Kind regards,
ict@etrovub.be

1.    Install and use OpenVPN

To access machines in the ETROVUB bubble (using an etrovub account), please use an OpenVPN connection.

In order to download the software, surf to https://vpn.etrovub.be and login with your etrovub credentials.

Download the latest software it offers to you and install.

Once done you can open it and connect with the same etrovub credentials (username only, no etrovub before or after), and click ok.

The connection establishes and shows:

2.    Generate RSA 2048-bit OpenSSH keys

When generating SSH keys, the private key should always be stored securely on your local machine,

while the public key is meant to be shared—simply copy it and send it to the other party.

You’ll also be asked whether to use a passphrase: this is optional.

It’s recommended to use a passphrase on laptops or shared systems for extra protection,

but you can skip it if you’re setting up automation or scripts where prompts would get in the way .

Here’s how to generate an RSA 2048 OpenSSH key pair on Windows, macOS, or Linux:

Windows (with PuTTYgen)        

• Open PuTTYgen
• Select:

• Type: RSA
• Bits: 2048

• Click Generate → move mouse
• (Optional) Set a Key passphrase
• Copy public key from ‘Public key for pasting into OpenSSH…’

• send to the other party
• Click Save private key → store as id_rsa.ppk locally

MacOS (Terminal)

ssh-keygen -t rsa -b 2048 -f ~/.ssh/id_rsa

• Hit Enter to skip passphrase for automation
• Show public key:
• cat ~/.ssh/id_rsa.pub → send to the other party
• Private key stored: ~/.ssh/id_rsa

Linux (Terminal)

Same as macOS:

ssh-keygen -t rsa -b 2048 -f ~/.ssh/id_rsa

• Skip passphrase if automating
• Show public key:
• cat ~/.ssh/id_rsa.pub → send to the other party
• Private key stored: ~/.ssh/id_rsa

3.    Connect to ETROFARM

login using ssh or equivalent

on: etroflock.etrovub.be (10.0.5.202)

using your

• credential login (without the @…)
• private key.

4.    Why the GPU Farm at ETRO?

At ETRO, advancing cutting-edge research hinges on the effective use of High-Performance Computing (HPC) resources. We recognize the need to bridge the gap between desktop computing limitations and the extensive power of HPC systems like Hydra. Our new dedicated GPU farm is designed to transform how users engage with computing resources—moving from power-hungry, noisy desktops in overheated rooms to a more flexible, efficient, and scalable solution.

This upgrade is not just about new hardware; it’s about reshaping user behavior and enhancing integration with VUB’s existing HPC platforms. By aligning the GPU farm with VUB’s systems, we ensure smoother project transitions and better compatibility with Hydra, allowing researchers to focus more on innovation rather than infrastructure.

The GPU farm will help ETRO achieve cost-efficiency, reduce our environmental footprint, and support research with greater flexibility and computational power—all without the high costs associated with commercial cloud solutions or the limitations of local desktops.

Available Resources

Hardware

We offer three nodes with the following configurations:

Partition	Node	CPU	GPU
FARM	ETROFARM	AMD EPYC 9124 (F. 25 M.17) 2 sockets 2x16x1= 32 (logical) CPU 377GB	4x Nvidia A100 (Ampere)   Nvidia Driver: 535.183.01   CUDA: 12.2   80GB
COOP	ETROCOOP01 ETROCOOP02	INTEL 13th Gen Intel(R) Core(TM) i9-13900 (F.6 M.183)   1 socket 1x8x2= 16 (logical) CPU   125GB	2x Nvidia GeForce RTX4090 (Ada Lovelace)   Nvidia Driver: 535.183.01   CUDA: 12.2 24GB

Additional Information:

• Data is accessible via OPENVPN.
• Network transfer rate: 30MB/s.
• Diskspace: 1TB workspace provided, with the possibility of creating a group workspace upon request. Diskspace is shared over a 1 Gbps network.

Software

We provide software built with Easybuild and managed with Lmod. While we strive to offer packages similar to those available on Hydra, our node architecture and resources are different. We aim to include the most common packages and provide configurations similar to those found at Hydra via Easybuild

Easyconfigs

Users can supply their own build files if necessary.
Note: Singularity containers (as used in Hydra) are not provided.

Priority

We use SLURM’s job accounting and fairshare system to manage resource allocation and prevent monopolization.
The fairshare score reflects cluster usage and helps prioritize jobs.

Training Resources

For SLURM functionality, see the theoretical guide at SLURM:
https://slurm.schedmd.com/.

Practical documentation closely aligned with our setup can be found at Hydra Documentation:
https://hpc.vub.be/docs/

Hydra also offers regular training sessions.

Two of ETRO’s postdocs, Angel for the project Equitable Oxymetry and Abel for the I-Healthy path project, have been selected for the MedTech accelerator: https://lifetech.brussels/en/medtech-accelerator-en/.

More info and pictures in these links:

• https://www.linkedin.com/posts/lifetech-brussels_medtechaccelerator2025-value-patients-activity-7305982232211476481-9Sh9?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAApXK0BavQNkWiiuuorRLD6_ShNleRksEs
• https://lifetech.brussels/en/medtech-accelerator-en/

Brussels, March 8, 2025 – On International Women’s Day, a day dedicated to celebrating women’s achievements and advocating for greater gender equality, the new WATS ambassador is announced: Lesley De Cruz. This inspiring professor at the Vrije Universiteit Brussel (VUB) and researcher at the Royal Meteorological Institute of Belgium (RMI) is committed to encouraging girls and women to pursue careers in STEM (science, technology, engineering, and mathematics). 

The WATS Award: Increasing Visibility for Women in STEM 

Since 2018, the Brussels-Capital Region and Innoviris have been presenting the biennial Women Award in Technology and Science (WATS). This award highlights Brussels-based women who not only excel in STEM but also inspire others to follow in their footsteps. The prize is awarded based on their potential as role models and their concrete plans to encourage girls and women to consider STEM studies or careers. 

Why Lesley De Cruz? 

Lesley De Cruz was selected by a jury of former laureates and young female STEM enthusiasts (aged 16-17) for her versatility as a scientist and her approachability as a role model. She also excels in her ability to communicate complex scientific topics in an accessible and engaging way. 

Her award-winning project introduces young people to artificial intelligence by guiding them to build a virtual city and examine its climate impact. This innovative approach makes scientific knowledge both relevant and tangible while addressing societal issues such as climate change. 

Making STEM Accessible for Everyone 

Lesley’s mission as a WATS ambassador is clear: “I want to show girls that STEM empowers you to make a positive impact on the world,” she says. “My message is especially aimed at girls who think, ‘Amazing, but that’s not for me.’ I want to inspire them to believe in themselves and show that anyone, regardless of background, can have a future in STEM.” 

With the accompanying €10,000 prize, Lesley plans to organize various activities in Brussels to actively guide and motivate girls. 

International Women’s Day: Celebrating Female Role Models 

On this International Women’s Day, we celebrate not only the achievements of women worldwide but also the role models who make a difference. Thanks to Lesley De Cruz and her efforts, we are building a future where STEM is a place for everyone, regardless of gender. 

On March 24th 2025 at 16:00, Selene De Sutter will defend their PhD entitled “SEGMENTATION OF GLIOBLASTOMA FROM MULTI-MODAL MEDICAL IMAGING – Towards revealing tumor infiltration”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Glioblastoma is the most aggressive type of primary brain tumor, and its treatment relies on accurately identifying the tumor’s boundaries – a process known as segmentation. However, manually outlining these areas is labor-intensive and can be susceptible to variability between experts.

Tumor boundaries are typically defined using magnetic resonance imaging (MRI), but glioblastoma cells often spread beyond what is visible on these scans, further complicating the segmentation process. These hidden, infiltrating cells are difficult to detect, often going untreated and contributing to tumor relapse. Detecting them is therefore crucial for improving treatment outcomes. Positron emission tomography (PET) with amino acid tracers may provide additional information about these infiltrating cells, but automated segmentation on these images remains underexplored.

This work investigates how image analysis and artificial intelligence techniques can enable accurate and automated segmentation of glioblastoma from medical images. The proposed contributions are initially focused on segmentation of the known tumor boundaries on MRI, while examining how different types of MRI scans contribute to this task. The focus then shifts towards segmentation using PET imaging, where we highlight its current limitations. Finally, we combine information from both MRI and PET scans to achieve a more comprehensive and accurate tumor segmentation, which may contribute to improved treatment planning and outcomes.

On February 20th 2025 at 16:00, Ángel Solé Morillo will defend their PhD entitled “M3-PPG: TOWARDS NOVEL (PERSONALIZED) PHOTOPLETHYSMOGRAPHY SYSTEMS THROUGH THE UNDERSTANING OF KEY INFLUENCING FACTORS”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Photoplethysmography (PPG) is a low-cost technique that allows for extracting physiological parameters, such as heart rate or blood oxygen saturation, through light interactions with the skin. PPG has been present in clinical practice as the technology behind pulse oximeters since the 1980s. With the proliferation of health wearables equipped with PPG sensors in the last 15 years and the advancement in PPG applications beyond pulse oximetry, a new perspective has arisen, with PPG having the potential to tackle some key societal and health issues of the 21st century.

Despite its widespread adoption, PPG remains susceptible to various factors that can compromise the accuracy of the physiological measurements. Understanding these influences individually can improve and expand the use and applications of PPG, ultimately enabling personalized health monitoring.

This research first proposes a theoretical framework, which describes key hardware and software improvements that can enable robust and personalized physiological monitoring using PPG technology. Next, an analysis of the impact of the instrumentation on the PPG signal is presented.

In addition, the impact of skin tone on the PPG signal is evaluated. Melanin, another PPG influencing factor whose content in the skin gives rise to different skin colors, is measured with two prototypes in a pilot study. This contributes to addressing the oxygen saturation overestimation in pulse oximeters for users with darker skin tones.

The final part integrates all previous research findings into a prototype designed for continuous vital sign monitoring at the chest, whose performance was validated through an initial pilot study with healthy participants. This work also analyses how regulations impact this prototype’s possible road to market as a medical device in Europe and the U.S.

On February 5th 2025 at 16:00, Raees Kizhakkumkara Muhamad will defend their PhD entitled “COMPRESSION STRATEGIES FOR DIGITAL HOLOGRAPHY”.

Everybody is invited to attend the presentation in room I.0.01 or online via this link.

Holographic techniques sample a 2D interference pattern produced by coherent light waves reflected or transmitted from the different objects in the 3D scene. As a display technology, it provides all necessary visual cues for perceiving the scene by the brain without causing mismatches between accommodation and vergence of the eyes. Non destructive imaging with high resolution for biomedical and industrial inspection also utilizes holographic principles. Holographic microscopes are realizable with optically simpler setups than regular microscopy, opening new pathways for computational microscopy. Utilizing more complex arrangements, such as holographic tomography, allows for reconstructing the 3D refractive index profile of transmissive objects, resolving even sub-cellular structures with visible light. It represents the culmination of humanity’s effort to record and represent light information.

However, sampling interference patterns for high-end displays or highresolution microscopy result in an extensive digital footprint. Historically, for many multimedia use cases, the data transmission bottleneck dictated the fidelity of the consumed content, and one can expect holography to be no different. Compression algorithms can help mitigate the data load, trading off more computation for an effective increase in transmission capacity. The algorithms must be tailored for holograms used in practice and exhibit a computational complexity appropriate for the use case, particularly on the decoder side. This thesis presents compression strategies for effectively tackling such use cases for holography.

The performance of conventional image compression tools on metrological hologram content is first studied. We provide a novel analysis of compression artifacts on the retrieved metrological data obtained for digital holographic microscopy and tomography. First-generation holographic displays are poised to use binary representation due to difficulties in modulating pixels at the sizes required by holography. Context-based compression is adequate for lossless and near-lossless compression of such data. Here, one extracts from a pixel the redundant information from previously decoded neighbouring pixels using a generalized Markovian model. Another context-based framework utilizing linear (autoregressive) models is used to design a highly scalable lossless compression scheme for non-binary holographic data. For broadcast scenarios, it is desirable to have a compression system that can support view-selective decoding to minimize the transmission of unutilized information. For this purpose, we propose a short-time Fourier transform (STFT) based codec, which slices the hologram into independently decoded, spatio-angular chunks. Given a target mean-squared error, the optimization techniques efficiently round down the signal in the STFT domain by application of adaptive quantizers. All these compression schemes are applicable for a single frame of hologram data and feature lightweight decoding architectures while surpassing compression performances achieved by any existing solution on most tested holograms. To compress holographic videos with arbitrary motion, we utilize a novel motion compensation algorithm that can predict rotational motion in conjunction with the above-mentioned STFT framework. The Markovian and STFT frameworks discussed in this work have been adopted as part of the first international hologram compression standard, JPEG Pleno — Part 5: Holography (ISO/IEC 21794-5)

ETRO, the Vrije Universiteit Brussel (VUB) and imec are proud to announce that Prof. Nikos Deligiannis has been awarded a prestigious ERC Consolidator Grant from the European Research Council to make a groundbreaking contribution to science and society.

Project: IONIAN: Reinventing Multiterminal Coding for Intelligent Machines
Budget: €1,999,404

Professor Deligiannis’ IONIAN project focuses on reinventing multiterminal coding, a crucial technology for efficient communication and collaboration between intelligent machines. With the explosive growth of data, such as video and point cloud streams, current storage and communication technologies are under pressure, undermining the ability of intelligent machines to cooperatively perceive their environment. This project develops a groundbreaking compression and communication approach based on interpretable and explainable AI that breaks the limits of traditional compression and cooperative perception techniques.

IONIAN combines classical theories, such as distributed source coding, with modern deep learning techniques and explainable AI, focusing on three innovative pillars:

1. Solving theoretical limitations in multiterminal coding.
2. Designing interpretable and efficient machine learning models.
3. Developing new methods to enhance model interpretation.

The goal of this project is to elevate the collaboration between intelligent systems, such as autonomous vehicles and mobile robots, to a higher level, with greater safety and trust as the result.

Remote editing with VS Code

Visual Studio Code is a widely-used, cross-platform Integrated Development Environment (IDE) that supports numerous programming languages and offers a vast array of extensions to enhance its functionality.

One notable extension enables development on remote machines via SSH, providing integrated access to a file explorer, terminal, and text editor on the configured remote system. This makes it a strong alternative to JupyterLab as a remote editor for e.g., the ETROFARM Slurm cluster.

Configuring Visual Studio Code extensions for working on ETROFARM Slurm cluster

Open the Extensions tab using the corresponding icon in the left toolbar.

Search for “SSH” and subsequently select the topmost “Remote – SSH” extension. Install this extension.

Depending on your programming language of choice, you might also be interested in extensions such as “Python”, “Python Debugger”, “Ruff” (a Python code linter) etc.

With the “Remote – SSH” extension installed, a new tab “Remote explorer” has been added to the left toolbar.

Add a new remote by pressing the + icon.

When asked for the SSH Connection Command:

ssh <username>@etroflock.etrovub.be

Secondly, it will ask where to store this information. This can be the default option.

The etroflock.etrovub.be remote has been created. Time to connect by triggering one of the 2 corresponding buttons.
During a brief instant the option will be displayed to edit the configuration. If you have missed it you can find this file via the gear button next to remote explorer – remote tunnels – ssh

Your config file should look like

Host etroflock.etrovub.be
  HostName etroflock.etrovub.be
  User jdoe
  IdentityFile C:\users\jdoe\.ssh\id_rsa

If you have no experience with encryption you can e.g. use a rsa 2048 type of key. Please make sure you are using a private key in openssh format.

Upon our first connection attempt, it requests the platform of the remote host, being the Slurm cluster’s login node we are connecting to. This is a Linux machine.

It will also ask to confirm the SSH public key credential of the server.

We are now connected to the remote server. This can be seen in the Remote explorer tab as well as in the left corner of the bottom toolbar.

Configuring the file explorer and text editor on the remote system

Time to open our file explorer via the “Explorer” tab in the left toolbar (Ctrl + Shift + E). Press the “Open Folder” button. It should by default suggest to open your home folder on the Slurm cluster (currently on /FARM/<username>).

If prompted, confirm that the remote server is (again) a Linux platform. Lastly, confirm that you trust the authors of the files in this folder as this is your own home folder.

Congratulations! Your remote file explorer and text editor on the Slurm cluster is now operational.

Remote terminal and running code on the remote system

A remote terminal can be opened using Terminal -> New Terminal in the top toolbar, or via the Ctrl + Shift + ` shortcut.

An interactive terminal session is opened on the remote host as if it was a PuTTY (or other) SSH session.

Running code remotely with VScode (ideally on Slurm compute nodes)

With the remote file explorer, remote text editor and remote terminal sessions available, it is a logical next step to focus on running our code remotely on the machine. Luckily, this is typically as straightforward as pressing the “Run file” button on the active (Python) file.

We can observe that the code has indeed been executed on the remote machine. However, the configured remote machine is ETROflock, ETRO’s Slurm cluster’s login server that is scarce in compute resources and lacks and GPU’s.

Running our code on a Slurm compute node is more complicated as it involves requesting a Slurm job. This is currently a manual process but we are investigating if this can be automated in VScode by using a custom launch script.

For now, there are two possibilities to run code on the Slurm cluster from within VScode:

A Slurm job can be requested that immediately runs the code until completion (or an error or timeout). This is the recommended default approach for running Slurm jobs.

E.g. the same test.py code is run as a Slurm job by using the srun command.

We observe that the first command is run on ETROFLOCK (the Slurm login node) because that is immediately launched on the remote system. The second command is scheduled as a Slurm job and is run on ETROFARM (a Slurm compute node).

The second possibility involves starting a Slurm job with an interactive shell. Once this interactive shell is running on a compute node, we can manually launch the desired code within that shell. This solution might be preferred when developing and testing the functionality of the code as a Slurm job (and potential queue) must only be requested once per session.

Launching a Slurm job with interactive shell is possible using the following srun command parameters:

srun –pty bash -i

E.g. the same test.py code is run after an interactive shell has been requested using the srun command and has started.

Observe that after requesting the slurm job with interactive terminal we change from “steffen@ETROFLOCK” to “steffen@ETROFARM”. This indicates our interactive shell is indeed running on a compute node instead of the login node.

Executing the test.py script from within this shell again confirms that the code is indeed run on the compute node with hostname ETROFARM.

When finished, one should use the “exit” command. This closes the interactive shell and terminates the Slurm job, thereby releasing the allocated resources for new jobs.

Filezilla is a free and open-source file transfer tool that can be used to exchange files with e.g., ETROFARM Slurm cluster. It requires some brief configuration in order to be able to connect using the SSH public/private key credentials. The required configuration steps are provided in the following tutorial.

Open the Site Manager, via the leftmost icon in the toolbar. Contrary to using the Quickconnect function which only supports connecting via a password, the Site Manager also allows configuring SFTP authentication via SSH keys.

Add a New site with the following configuration:

1. Protocol: SFTP – SSH File Transfer Protocol
2. Port: Optional field, can remain empty (default SFTP value: 22)
3. Host: etroflock.etrovub.be  (IP address: 10.0.5.202)
4. Logon Type: Key file
5. User: Username of your slurm account, typically your ETRO username
6. Key file: Path to your private SSH key used for Slurm
   Note: When browsing, you might need to look for “All files” as there might be no file extension. For UNIX-style systems (Linux / MAC) this is typically found in the user’s home / .ssh directory which is a directory that might be hidden.

Connect to this SFTP server and trust the etroflock’s public SSH key.

All future FileZilla SFTP connections to ETROflock can be easily launched from the Site manager tab.

On November 15th 2024 at 10:00, Eden Teshome Hunde will defend their PhD entitled “CROSS-LAYER DESIGN, IMPLEMENTATION AND EVALUATION OF IPV6 MULTICAST FOR RADIO DUTY CYCLED WIRELESS SENSOR AND ACTUATOR NETWORKS”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Wireless sensor and actuator networks (WSANs), consist of small-sized constrained sensing and/or actuating devices which can run an Internet Protocol version 6 (IPv6) network stack. In such WSANs, there is often the need for sending the same message to several devices in the WSAN, such as for applications for controlling all the lights in a building or turning on/off all the air conditioners in the room etc. For doing so, IP multicast can be an efficient solution.

In this work, we study Bidirectional Multicast RPL Forwarding (BMRF) as this protocol relies on forwarding tables put in place by the well-known Routing Protocol for Low Power and Lossy Networks (RPL) and allows to combine the best ideas of existing multicast protocols. Through RPL, a routing tree towards the sink is installed for multihop routing from node to sink, and the nodes’ forwarding tables will also contain entries for reaching destinations in downward direction.

For downward forwarding IPv6 multicast packets, two methods exist. One is via link layer (LL), broadcasting a frame containing the IPv6 multicast packet. The other is to send several LL unicast frames containing that packet. BMRF allows a node to choose between these two methods. The best option will depend on the presence of a radio duty cycling (RDC) protocol. RDC is part of the medium access control (MAC) layer and puts the radio to sleep when no communication is needed. We investigate the influence of MAC/RDC protocols on BMRF’s performance.

We evaluate the performance of BMRF on non-synchronized WSANs that use Carrier Sense Multiple Access (CSMA) as MAC and ContikiMAC as RDC. We demonstrate that LL unicast outperforms LL broadcast in terms of packet delivery ratio (PDR), delay, and energy consumption in many settings.

We investigate the performance of BMRF on WSANs with synchronous MAC and RDC based on Time Slotted Channel hopping (TSCH). This is more challenging, as TSCH needs a schedule to tell which action must happen in each timeslot. The actions can be to send or to listen on a given channel or to be idle. Idleness allows the radio to switch OFF, providing RDC. The schedule is not part of the standard and must be proposed by the system designer. An elegant autonomous scheduling method called Orchestra is available to accommodate traffic in a RPL tree. We extend Orchestra with a novel scheduling rule for supporting LL downwards forwarding through LL broadcast. Comparing LL unicast with LL broadcast forwarding teaches us that LL unicast outperforms LL broadcast in terms of packet delivery ratio (PDR), but the latter can be beneficial to certain applications, especially those sensitive to delay.

Before conducting the two previous evaluation studies, we investigate the performance of simple convergecast traffic while considering ContikiMAC and TSCH with Orchestra under RPL on the real dual Zolertia Firefly Motes (one is observed and other one is observing mote). This study served two purposes; it reminds the reader of the characteristics of those protocols and allowed to fine-tune the dual motes.

We also contributed by adapting the Orchestra to bursty convergecast traffic. Simulation results demonstrate that the new scheduler slightly improves PDR and reduces delay compared to state-of-the-art solutions.

“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 April 25th 2025 at 16:00, Ayman Morsy will defend their PhD entitled “A NOVEL APPROACH TO DEPTH-SENSE IMAGING USING CORRELATION-ASSISTED DIRECT TIME-OF-FLIGHT”.

Everybody is invited to attend the presentation in room I.0.03 or online via this link.

Time-of-flight (ToF) imaging has emerged as a vital technology in machine vision and sensing, expanding into applications such as augmented and virtual reality, gaming, robotics, autonomous driving, autofocus, and facial recognition on smartphones and laptops. ToF technology determines the distance to an object within the detection range by emitting a light source and measuring the time it takes to return. This round-trip time determines the object’s distance, with different sensing technologies employing distinct methods to determine this time.

For ToF applications, developing sensors with high image resolution, low power consumption, and the ability to function reliably in high ambient light conditions is desirable. This dissertation presents the development of a novel single-photon avalanche diode (SPAD)-based pixel called Correlation-Assisted Direct Time-of-Flight (CA-dToF), designed for in-pixel ambient light suppression and characterized by low power consumption and a scalable pixel structure. The CA-dToF pixel uses a laser pulse correlated with two orthogonal sinusoidal signals as input to two switched capacitor channels, which average out detected ambient light while accumulating the laser pulse round-trip time.

To gain insights into CA-dToF pixel operation, both Python simulation and analytical modeling were developed. Two generations of the CA-dToF pixel were developed and characterized, with the second-generation pixel achieving the first operational performance under high ambient light conditions. The two-generation CA-dToF pixel was tested under various lighting conditions and pixel design variations. Additionally, noise sources within the pixel implementation were analyzed, and potential solutions were proposed.

Dear [Newcomer],

Welcome to the ETRO.RDI LARGE research group! We are excited to have you on board.

Your colleagues will be happy to assist you as you settle in. However, they may not have all the answers—that’s where we, the central support team of ETRO.RDI, come in.

At ETRO.RDI, we believe that getting to know each other makes everything smoother. Collaboration becomes more efficient, and let’s be honest—it’s just more fun!

We understand that integrating into a new team takes some effort, but the sooner you start, the quicker you’ll feel at home.

Let’s avoid a bumpy start; please complete the following steps smoothly.

Immediately

✅ Introduce yourself – Share three short bullet points with personal details you’d like to share with the ETRO team. These will be used for your introduction via ETRO.TV, one of our internal communication platforms.

Once You Receive Your ETRO Credentials

✅ Log in to the intranet – Visit www.etrovub.be and click the icon in the top-right corner to log in. Take some time to explore the platform.

✅ Complete your personal ETRO.RDI webpage – Log in at https://www.etrovub.be/wp-admin/ and update your profile using your etrovub username and password (without adding “etrovub\”).

✅ Schedule your web picture appointment – Once you are in the country, contact Luc van Kempen to arrange your professional photo session. This ensures a uniform presentation of yourself and your research topic on our website and internal communication channels.

By completing these steps, you’ll make it easier to connect with colleagues. The first time you meet someone, chances are you’ll already know a little about each other—perhaps even sharing a common interest you’ve mentioned earlier.

---

If you ever feel unsure or need assistance, don’t hesitate to reach out (see the contact details below).
We’re here to support you! Sometimes, we’ll help you directly, and other times, we may refer you to the intranet or connect you with a colleague who has encountered a similar challenge.

📧 For research-related or administrative questions:

Karin De Bruyn, ETRO Research Coordinator – Karin.Debruyn@vub.be

💻 For IT-related support:
ICT@etrovub.be – We’ll assist you directly or forward your request to the central helpdesk if needed.

🔧 For technical, purchase or access issues:

Luc van Kempen, ETRO System Manager –  Luc.van.Kempen@vub.be We look forward to meeting you!

Dear Future User of ETROFARM,

To process your access request, please provide the following information:

• Your Last Name
• Your First Name
• A valid email address for communication (preferably your VUB email; using it for the request is ideal)
• Your Principal Investigator’s (PI) name (adding them in CC is ideal)
• Your public SSH key in OpenSSH format (2), which will be used to connect to ETROFARM

Steps to follow:

1. Send your access request to ICT@etrovub.be
2. You will receive an initial email containing your ETROVUB credentials, allowing you to:
3. Install OpenVPN (1) (if not already installed) and connect to the ETRO Bubble
4. Receive a second email confirming your access to ETROFARM
5. Connect to ETROFARM (3)
6. Read the provided documentation (4)

This is our standard procedure. If you’ve already provided some of the requested information, there’s no need to resend it.

Kind regards,
ict@etrovub.be

1.    Install and use OpenVPN

To access machines in the ETROVUB bubble (using an etrovub account), please use an OpenVPN connection.

In order to download the software, surf to https://vpn.etrovub.be and login with your etrovub credentials.

Download the latest software it offers to you and install.

Once done you can open it and connect with the same etrovub credentials (username only, no etrovub before or after), and click ok.

The connection establishes and shows:

2.    Generate RSA 2048-bit OpenSSH keys

When generating SSH keys, the private key should always be stored securely on your local machine,

while the public key is meant to be shared—simply copy it and send it to the other party.

You’ll also be asked whether to use a passphrase: this is optional.

It’s recommended to use a passphrase on laptops or shared systems for extra protection,

but you can skip it if you’re setting up automation or scripts where prompts would get in the way .

Here’s how to generate an RSA 2048 OpenSSH key pair on Windows, macOS, or Linux:

Windows (with PuTTYgen)        

• Open PuTTYgen
• Select:

• Type: RSA
• Bits: 2048

• Click Generate → move mouse
• (Optional) Set a Key passphrase
• Copy public key from ‘Public key for pasting into OpenSSH…’

• send to the other party
• Click Save private key → store as id_rsa.ppk locally

MacOS (Terminal)

ssh-keygen -t rsa -b 2048 -f ~/.ssh/id_rsa

• Hit Enter to skip passphrase for automation
• Show public key:
• cat ~/.ssh/id_rsa.pub → send to the other party
• Private key stored: ~/.ssh/id_rsa

Linux (Terminal)

Same as macOS:

ssh-keygen -t rsa -b 2048 -f ~/.ssh/id_rsa

• Skip passphrase if automating
• Show public key:
• cat ~/.ssh/id_rsa.pub → send to the other party
• Private key stored: ~/.ssh/id_rsa

3.    Connect to ETROFARM

login using ssh or equivalent

on: etroflock.etrovub.be (10.0.5.202)

using your

• credential login (without the @…)
• private key.

4.    Why the GPU Farm at ETRO?

At ETRO, advancing cutting-edge research hinges on the effective use of High-Performance Computing (HPC) resources. We recognize the need to bridge the gap between desktop computing limitations and the extensive power of HPC systems like Hydra. Our new dedicated GPU farm is designed to transform how users engage with computing resources—moving from power-hungry, noisy desktops in overheated rooms to a more flexible, efficient, and scalable solution.

This upgrade is not just about new hardware; it’s about reshaping user behavior and enhancing integration with VUB’s existing HPC platforms. By aligning the GPU farm with VUB’s systems, we ensure smoother project transitions and better compatibility with Hydra, allowing researchers to focus more on innovation rather than infrastructure.

The GPU farm will help ETRO achieve cost-efficiency, reduce our environmental footprint, and support research with greater flexibility and computational power—all without the high costs associated with commercial cloud solutions or the limitations of local desktops.

Available Resources

Hardware

We offer three nodes with the following configurations:

Partition	Node	CPU	GPU
FARM	ETROFARM	AMD EPYC 9124 (F. 25 M.17) 2 sockets 2x16x1= 32 (logical) CPU 377GB	4x Nvidia A100 (Ampere)   Nvidia Driver: 535.183.01   CUDA: 12.2   80GB
COOP	ETROCOOP01 ETROCOOP02	INTEL 13th Gen Intel(R) Core(TM) i9-13900 (F.6 M.183)   1 socket 1x8x2= 16 (logical) CPU   125GB	2x Nvidia GeForce RTX4090 (Ada Lovelace)   Nvidia Driver: 535.183.01   CUDA: 12.2 24GB

Additional Information:

• Data is accessible via OPENVPN.
• Network transfer rate: 30MB/s.
• Diskspace: 1TB workspace provided, with the possibility of creating a group workspace upon request. Diskspace is shared over a 1 Gbps network.

Software

We provide software built with Easybuild and managed with Lmod. While we strive to offer packages similar to those available on Hydra, our node architecture and resources are different. We aim to include the most common packages and provide configurations similar to those found at Hydra via Easybuild

Easyconfigs

Users can supply their own build files if necessary.
Note: Singularity containers (as used in Hydra) are not provided.

Priority

We use SLURM’s job accounting and fairshare system to manage resource allocation and prevent monopolization.
The fairshare score reflects cluster usage and helps prioritize jobs.

Training Resources

For SLURM functionality, see the theoretical guide at SLURM:
https://slurm.schedmd.com/.

Practical documentation closely aligned with our setup can be found at Hydra Documentation:
https://hpc.vub.be/docs/

Hydra also offers regular training sessions.

Two of ETRO’s postdocs, Angel for the project Equitable Oxymetry and Abel for the I-Healthy path project, have been selected for the MedTech accelerator: https://lifetech.brussels/en/medtech-accelerator-en/.

More info and pictures in these links:

• https://www.linkedin.com/posts/lifetech-brussels_medtechaccelerator2025-value-patients-activity-7305982232211476481-9Sh9?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAApXK0BavQNkWiiuuorRLD6_ShNleRksEs
• https://lifetech.brussels/en/medtech-accelerator-en/

Brussels, March 8, 2025 – On International Women’s Day, a day dedicated to celebrating women’s achievements and advocating for greater gender equality, the new WATS ambassador is announced: Lesley De Cruz. This inspiring professor at the Vrije Universiteit Brussel (VUB) and researcher at the Royal Meteorological Institute of Belgium (RMI) is committed to encouraging girls and women to pursue careers in STEM (science, technology, engineering, and mathematics). 

The WATS Award: Increasing Visibility for Women in STEM 

Since 2018, the Brussels-Capital Region and Innoviris have been presenting the biennial Women Award in Technology and Science (WATS). This award highlights Brussels-based women who not only excel in STEM but also inspire others to follow in their footsteps. The prize is awarded based on their potential as role models and their concrete plans to encourage girls and women to consider STEM studies or careers. 

Why Lesley De Cruz? 

Lesley De Cruz was selected by a jury of former laureates and young female STEM enthusiasts (aged 16-17) for her versatility as a scientist and her approachability as a role model. She also excels in her ability to communicate complex scientific topics in an accessible and engaging way. 

Her award-winning project introduces young people to artificial intelligence by guiding them to build a virtual city and examine its climate impact. This innovative approach makes scientific knowledge both relevant and tangible while addressing societal issues such as climate change. 

Making STEM Accessible for Everyone 

Lesley’s mission as a WATS ambassador is clear: “I want to show girls that STEM empowers you to make a positive impact on the world,” she says. “My message is especially aimed at girls who think, ‘Amazing, but that’s not for me.’ I want to inspire them to believe in themselves and show that anyone, regardless of background, can have a future in STEM.” 

With the accompanying €10,000 prize, Lesley plans to organize various activities in Brussels to actively guide and motivate girls. 

International Women’s Day: Celebrating Female Role Models 

On this International Women’s Day, we celebrate not only the achievements of women worldwide but also the role models who make a difference. Thanks to Lesley De Cruz and her efforts, we are building a future where STEM is a place for everyone, regardless of gender. 

On March 24th 2025 at 16:00, Selene De Sutter will defend their PhD entitled “SEGMENTATION OF GLIOBLASTOMA FROM MULTI-MODAL MEDICAL IMAGING – Towards revealing tumor infiltration”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Glioblastoma is the most aggressive type of primary brain tumor, and its treatment relies on accurately identifying the tumor’s boundaries – a process known as segmentation. However, manually outlining these areas is labor-intensive and can be susceptible to variability between experts.

Tumor boundaries are typically defined using magnetic resonance imaging (MRI), but glioblastoma cells often spread beyond what is visible on these scans, further complicating the segmentation process. These hidden, infiltrating cells are difficult to detect, often going untreated and contributing to tumor relapse. Detecting them is therefore crucial for improving treatment outcomes. Positron emission tomography (PET) with amino acid tracers may provide additional information about these infiltrating cells, but automated segmentation on these images remains underexplored.

This work investigates how image analysis and artificial intelligence techniques can enable accurate and automated segmentation of glioblastoma from medical images. The proposed contributions are initially focused on segmentation of the known tumor boundaries on MRI, while examining how different types of MRI scans contribute to this task. The focus then shifts towards segmentation using PET imaging, where we highlight its current limitations. Finally, we combine information from both MRI and PET scans to achieve a more comprehensive and accurate tumor segmentation, which may contribute to improved treatment planning and outcomes.

On February 20th 2025 at 16:00, Ángel Solé Morillo will defend their PhD entitled “M3-PPG: TOWARDS NOVEL (PERSONALIZED) PHOTOPLETHYSMOGRAPHY SYSTEMS THROUGH THE UNDERSTANING OF KEY INFLUENCING FACTORS”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Photoplethysmography (PPG) is a low-cost technique that allows for extracting physiological parameters, such as heart rate or blood oxygen saturation, through light interactions with the skin. PPG has been present in clinical practice as the technology behind pulse oximeters since the 1980s. With the proliferation of health wearables equipped with PPG sensors in the last 15 years and the advancement in PPG applications beyond pulse oximetry, a new perspective has arisen, with PPG having the potential to tackle some key societal and health issues of the 21st century.

Despite its widespread adoption, PPG remains susceptible to various factors that can compromise the accuracy of the physiological measurements. Understanding these influences individually can improve and expand the use and applications of PPG, ultimately enabling personalized health monitoring.

This research first proposes a theoretical framework, which describes key hardware and software improvements that can enable robust and personalized physiological monitoring using PPG technology. Next, an analysis of the impact of the instrumentation on the PPG signal is presented.

In addition, the impact of skin tone on the PPG signal is evaluated. Melanin, another PPG influencing factor whose content in the skin gives rise to different skin colors, is measured with two prototypes in a pilot study. This contributes to addressing the oxygen saturation overestimation in pulse oximeters for users with darker skin tones.

The final part integrates all previous research findings into a prototype designed for continuous vital sign monitoring at the chest, whose performance was validated through an initial pilot study with healthy participants. This work also analyses how regulations impact this prototype’s possible road to market as a medical device in Europe and the U.S.

On February 5th 2025 at 16:00, Raees Kizhakkumkara Muhamad will defend their PhD entitled “COMPRESSION STRATEGIES FOR DIGITAL HOLOGRAPHY”.

Everybody is invited to attend the presentation in room I.0.01 or online via this link.

Holographic techniques sample a 2D interference pattern produced by coherent light waves reflected or transmitted from the different objects in the 3D scene. As a display technology, it provides all necessary visual cues for perceiving the scene by the brain without causing mismatches between accommodation and vergence of the eyes. Non destructive imaging with high resolution for biomedical and industrial inspection also utilizes holographic principles. Holographic microscopes are realizable with optically simpler setups than regular microscopy, opening new pathways for computational microscopy. Utilizing more complex arrangements, such as holographic tomography, allows for reconstructing the 3D refractive index profile of transmissive objects, resolving even sub-cellular structures with visible light. It represents the culmination of humanity’s effort to record and represent light information.

However, sampling interference patterns for high-end displays or highresolution microscopy result in an extensive digital footprint. Historically, for many multimedia use cases, the data transmission bottleneck dictated the fidelity of the consumed content, and one can expect holography to be no different. Compression algorithms can help mitigate the data load, trading off more computation for an effective increase in transmission capacity. The algorithms must be tailored for holograms used in practice and exhibit a computational complexity appropriate for the use case, particularly on the decoder side. This thesis presents compression strategies for effectively tackling such use cases for holography.

The performance of conventional image compression tools on metrological hologram content is first studied. We provide a novel analysis of compression artifacts on the retrieved metrological data obtained for digital holographic microscopy and tomography. First-generation holographic displays are poised to use binary representation due to difficulties in modulating pixels at the sizes required by holography. Context-based compression is adequate for lossless and near-lossless compression of such data. Here, one extracts from a pixel the redundant information from previously decoded neighbouring pixels using a generalized Markovian model. Another context-based framework utilizing linear (autoregressive) models is used to design a highly scalable lossless compression scheme for non-binary holographic data. For broadcast scenarios, it is desirable to have a compression system that can support view-selective decoding to minimize the transmission of unutilized information. For this purpose, we propose a short-time Fourier transform (STFT) based codec, which slices the hologram into independently decoded, spatio-angular chunks. Given a target mean-squared error, the optimization techniques efficiently round down the signal in the STFT domain by application of adaptive quantizers. All these compression schemes are applicable for a single frame of hologram data and feature lightweight decoding architectures while surpassing compression performances achieved by any existing solution on most tested holograms. To compress holographic videos with arbitrary motion, we utilize a novel motion compensation algorithm that can predict rotational motion in conjunction with the above-mentioned STFT framework. The Markovian and STFT frameworks discussed in this work have been adopted as part of the first international hologram compression standard, JPEG Pleno — Part 5: Holography (ISO/IEC 21794-5)

ETRO, the Vrije Universiteit Brussel (VUB) and imec are proud to announce that Prof. Nikos Deligiannis has been awarded a prestigious ERC Consolidator Grant from the European Research Council to make a groundbreaking contribution to science and society.

Project: IONIAN: Reinventing Multiterminal Coding for Intelligent Machines
Budget: €1,999,404

Professor Deligiannis’ IONIAN project focuses on reinventing multiterminal coding, a crucial technology for efficient communication and collaboration between intelligent machines. With the explosive growth of data, such as video and point cloud streams, current storage and communication technologies are under pressure, undermining the ability of intelligent machines to cooperatively perceive their environment. This project develops a groundbreaking compression and communication approach based on interpretable and explainable AI that breaks the limits of traditional compression and cooperative perception techniques.

IONIAN combines classical theories, such as distributed source coding, with modern deep learning techniques and explainable AI, focusing on three innovative pillars:

1. Solving theoretical limitations in multiterminal coding.
2. Designing interpretable and efficient machine learning models.
3. Developing new methods to enhance model interpretation.

The goal of this project is to elevate the collaboration between intelligent systems, such as autonomous vehicles and mobile robots, to a higher level, with greater safety and trust as the result.

Remote editing with VS Code

Visual Studio Code is a widely-used, cross-platform Integrated Development Environment (IDE) that supports numerous programming languages and offers a vast array of extensions to enhance its functionality.

One notable extension enables development on remote machines via SSH, providing integrated access to a file explorer, terminal, and text editor on the configured remote system. This makes it a strong alternative to JupyterLab as a remote editor for e.g., the ETROFARM Slurm cluster.

Configuring Visual Studio Code extensions for working on ETROFARM Slurm cluster

Open the Extensions tab using the corresponding icon in the left toolbar.

Search for “SSH” and subsequently select the topmost “Remote – SSH” extension. Install this extension.

Depending on your programming language of choice, you might also be interested in extensions such as “Python”, “Python Debugger”, “Ruff” (a Python code linter) etc.

With the “Remote – SSH” extension installed, a new tab “Remote explorer” has been added to the left toolbar.

Add a new remote by pressing the + icon.

When asked for the SSH Connection Command:

ssh <username>@etroflock.etrovub.be

Secondly, it will ask where to store this information. This can be the default option.

The etroflock.etrovub.be remote has been created. Time to connect by triggering one of the 2 corresponding buttons.
During a brief instant the option will be displayed to edit the configuration. If you have missed it you can find this file via the gear button next to remote explorer – remote tunnels – ssh

Your config file should look like

Host etroflock.etrovub.be
  HostName etroflock.etrovub.be
  User jdoe
  IdentityFile C:\users\jdoe\.ssh\id_rsa

If you have no experience with encryption you can e.g. use a rsa 2048 type of key. Please make sure you are using a private key in openssh format.

Upon our first connection attempt, it requests the platform of the remote host, being the Slurm cluster’s login node we are connecting to. This is a Linux machine.

It will also ask to confirm the SSH public key credential of the server.

We are now connected to the remote server. This can be seen in the Remote explorer tab as well as in the left corner of the bottom toolbar.

Configuring the file explorer and text editor on the remote system

Time to open our file explorer via the “Explorer” tab in the left toolbar (Ctrl + Shift + E). Press the “Open Folder” button. It should by default suggest to open your home folder on the Slurm cluster (currently on /FARM/<username>).

If prompted, confirm that the remote server is (again) a Linux platform. Lastly, confirm that you trust the authors of the files in this folder as this is your own home folder.

Congratulations! Your remote file explorer and text editor on the Slurm cluster is now operational.

Remote terminal and running code on the remote system

A remote terminal can be opened using Terminal -> New Terminal in the top toolbar, or via the Ctrl + Shift + ` shortcut.

An interactive terminal session is opened on the remote host as if it was a PuTTY (or other) SSH session.

Running code remotely with VScode (ideally on Slurm compute nodes)

With the remote file explorer, remote text editor and remote terminal sessions available, it is a logical next step to focus on running our code remotely on the machine. Luckily, this is typically as straightforward as pressing the “Run file” button on the active (Python) file.

We can observe that the code has indeed been executed on the remote machine. However, the configured remote machine is ETROflock, ETRO’s Slurm cluster’s login server that is scarce in compute resources and lacks and GPU’s.

Running our code on a Slurm compute node is more complicated as it involves requesting a Slurm job. This is currently a manual process but we are investigating if this can be automated in VScode by using a custom launch script.

For now, there are two possibilities to run code on the Slurm cluster from within VScode:

A Slurm job can be requested that immediately runs the code until completion (or an error or timeout). This is the recommended default approach for running Slurm jobs.

E.g. the same test.py code is run as a Slurm job by using the srun command.

We observe that the first command is run on ETROFLOCK (the Slurm login node) because that is immediately launched on the remote system. The second command is scheduled as a Slurm job and is run on ETROFARM (a Slurm compute node).

The second possibility involves starting a Slurm job with an interactive shell. Once this interactive shell is running on a compute node, we can manually launch the desired code within that shell. This solution might be preferred when developing and testing the functionality of the code as a Slurm job (and potential queue) must only be requested once per session.

Launching a Slurm job with interactive shell is possible using the following srun command parameters:

srun –pty bash -i

E.g. the same test.py code is run after an interactive shell has been requested using the srun command and has started.

Observe that after requesting the slurm job with interactive terminal we change from “steffen@ETROFLOCK” to “steffen@ETROFARM”. This indicates our interactive shell is indeed running on a compute node instead of the login node.

Executing the test.py script from within this shell again confirms that the code is indeed run on the compute node with hostname ETROFARM.

When finished, one should use the “exit” command. This closes the interactive shell and terminates the Slurm job, thereby releasing the allocated resources for new jobs.

Filezilla is a free and open-source file transfer tool that can be used to exchange files with e.g., ETROFARM Slurm cluster. It requires some brief configuration in order to be able to connect using the SSH public/private key credentials. The required configuration steps are provided in the following tutorial.

Open the Site Manager, via the leftmost icon in the toolbar. Contrary to using the Quickconnect function which only supports connecting via a password, the Site Manager also allows configuring SFTP authentication via SSH keys.

Add a New site with the following configuration:

1. Protocol: SFTP – SSH File Transfer Protocol
2. Port: Optional field, can remain empty (default SFTP value: 22)
3. Host: etroflock.etrovub.be  (IP address: 10.0.5.202)
4. Logon Type: Key file
5. User: Username of your slurm account, typically your ETRO username
6. Key file: Path to your private SSH key used for Slurm
   Note: When browsing, you might need to look for “All files” as there might be no file extension. For UNIX-style systems (Linux / MAC) this is typically found in the user’s home / .ssh directory which is a directory that might be hidden.

Connect to this SFTP server and trust the etroflock’s public SSH key.

All future FileZilla SFTP connections to ETROflock can be easily launched from the Site manager tab.

On November 15th 2024 at 10:00, Eden Teshome Hunde will defend their PhD entitled “CROSS-LAYER DESIGN, IMPLEMENTATION AND EVALUATION OF IPV6 MULTICAST FOR RADIO DUTY CYCLED WIRELESS SENSOR AND ACTUATOR NETWORKS”.

Everybody is invited to attend the presentation in room D.2.01 or online via this link.

Wireless sensor and actuator networks (WSANs), consist of small-sized constrained sensing and/or actuating devices which can run an Internet Protocol version 6 (IPv6) network stack. In such WSANs, there is often the need for sending the same message to several devices in the WSAN, such as for applications for controlling all the lights in a building or turning on/off all the air conditioners in the room etc. For doing so, IP multicast can be an efficient solution.

In this work, we study Bidirectional Multicast RPL Forwarding (BMRF) as this protocol relies on forwarding tables put in place by the well-known Routing Protocol for Low Power and Lossy Networks (RPL) and allows to combine the best ideas of existing multicast protocols. Through RPL, a routing tree towards the sink is installed for multihop routing from node to sink, and the nodes’ forwarding tables will also contain entries for reaching destinations in downward direction.

For downward forwarding IPv6 multicast packets, two methods exist. One is via link layer (LL), broadcasting a frame containing the IPv6 multicast packet. The other is to send several LL unicast frames containing that packet. BMRF allows a node to choose between these two methods. The best option will depend on the presence of a radio duty cycling (RDC) protocol. RDC is part of the medium access control (MAC) layer and puts the radio to sleep when no communication is needed. We investigate the influence of MAC/RDC protocols on BMRF’s performance.

We evaluate the performance of BMRF on non-synchronized WSANs that use Carrier Sense Multiple Access (CSMA) as MAC and ContikiMAC as RDC. We demonstrate that LL unicast outperforms LL broadcast in terms of packet delivery ratio (PDR), delay, and energy consumption in many settings.

We investigate the performance of BMRF on WSANs with synchronous MAC and RDC based on Time Slotted Channel hopping (TSCH). This is more challenging, as TSCH needs a schedule to tell which action must happen in each timeslot. The actions can be to send or to listen on a given channel or to be idle. Idleness allows the radio to switch OFF, providing RDC. The schedule is not part of the standard and must be proposed by the system designer. An elegant autonomous scheduling method called Orchestra is available to accommodate traffic in a RPL tree. We extend Orchestra with a novel scheduling rule for supporting LL downwards forwarding through LL broadcast. Comparing LL unicast with LL broadcast forwarding teaches us that LL unicast outperforms LL broadcast in terms of packet delivery ratio (PDR), but the latter can be beneficial to certain applications, especially those sensitive to delay.

Before conducting the two previous evaluation studies, we investigate the performance of simple convergecast traffic while considering ContikiMAC and TSCH with Orchestra under RPL on the real dual Zolertia Firefly Motes (one is observed and other one is observing mote). This study served two purposes; it reminds the reader of the characteristics of those protocols and allowed to fine-tune the dual motes.

We also contributed by adapting the Orchestra to bursty convergecast traffic. Simulation results demonstrate that the new scheduler slightly improves PDR and reduces delay compared to state-of-the-art solutions.