Welcome to the Swedish symposia on deep learning and image analysis!


This year's SSDL and SSBA symposia will be held online as back-to-back events. They replace SSDL/SSBA 2020, which had to be cancelled due to the pandemic. SSDL/SSBA 2021 thereby marks the continuation of a 38-year-old tradition of gathering researchers in the field of image analysis and pattern recognition who are active in Sweden to share their latest findings and socialize in an annual national forum.

SSDL is an important Swedish forum for leading research groups in industry and academia to meet and discuss the latest trends and developments in deep learning and related areas. SSDL 2021 will feature invited talks by leading researchers in deep learning as well as oral and poster presentations of submitted papers and abstracts.



The SSBA symposium is the premier Swedish event where researchers, industrial professionals and students gather to learn about the recent developments in the areas of image processing, computer vision, pattern recognition and related fields. SSBA 2021 features keynote speakers and oral presentations and posters of submitted papers.

SSDL/SSBA 2021 is organized by the Computer vision group at LTH Faculty of Engineering at Lund University, and Swedish Society for Automated Image Analysis (SSBA).

Important dates

Deadlines
SSBA paper submission deadline 1 March, 2021
SSDL abstract/paper submission deadline 1 March, 2021

Registration deadline

4 March, 2021




Conference dates
Deep Learning Symposium (SSDL) 15 March 13.00-17:00
Image Analysis Symposium (SSBA) 16-17 March 13.00-17:00

Registration

Open for registration!

Registration is available via ssdlssbaregister2021@ssba.org.se

Participation in SSDL/SSBA 2021 is FREE OF CHARGE. Please notice that the corresponding author of a contributed paper (SSDL or SSBA) or an extended abstract (SSDL only) is automatically registered to the symposia.
Everybody else will have to register by sending an email to ssdlssbaregister2021@ssba.org.se.

Registration deadline is 4 March, 2021

Submission

Open for submissions!

We welcome extended abstracts or regular papers for the SSDL symposium and regular papers for the SSBA symposium. We explicitly allow you to submit material that has already been published or is going to be submitted elsewhere. The content will only be available to symposium attendees.

See the detailed instructions below.

Submission deadline is 1 March, 2021



SSDL submissions


Contributions may be shorter, extended abstracts, or longer, a paper. You are free to choose the format you like, in particular there is no upper limit on the number of pages, but the contribution must be submitted as a PDF file. Moreover, the name of the PDF file should be in the format FamilynameGivenname.pdf, where Givenname and Familyname refer to the name of the corresponding author. Submit your abstract via email to ssdlabstract2021@ssba.org.se. Titles and authors of the contributions will be listed in the program of SSDL on this website. If your submission is based on a paper that is currently under review, make sure to use a different title to preserve anonymity.
INSTRUCTIONS ABOUT PRESENTATION FORMAT - POSTER or ORAL PRESENTATION - WILL BE GIVEN SOON.

Program

  • Monday, 15 March

    13:00 - 13:15 Welcome address
    13:15 - 14:15 Keynote 1: Robert Jenssen, Norges arktiske universitet, Tromsø Industrial and basic deep learning research for computer vision with limited labels
    14:15 - 16:00 Poster Session and mingle in Gather
    16:00 - 17:00 Keynote 2: Martin Danelljan, ETH Zürich Deep Visual Reasoning with Optimization-based Network Modules
    17:00 - 17:30 Mingle in Gather




  • Tuesday, March 16

    12:45 - 13:45 Keynote 3: Mattias Ohlsson, , Lunds universitet och Högskolan i Halmstad Deep learning in information-driven healthcare: Applications and model development
    13:45 - 14:45 Contributed talks - Session 1 (single track, 3 talks)
    14:45 - 15:00 Break and mingle in Gather
    15:00 - 16:00 Contributed talks - Session 2 (single track, 3 talks)
    16:00 - 17:00 Keynote 4: Mattias P Heinrich, Lübeck How to Learn 3D Lung Image Registration? Geometric Learning with Weak Supervision
    17:00 - 17:15 Award session i. Best Swedish Thesis in Image Analysis 2019-2020
    ii. Presentation of host for the 2022 SSDL/SSBA
    17:15 - 18:15 Annual meeting of the Swedish Society of Image Analysis (SSBA)



  • Wednesday, March 17

    12:30 - 13:50 Contributed talks, Session 3 (single track, 4 talks)
    13:50 - 14:00 Short break
    14:00 - 15:20 Contributed talks, Session 4 (single track, 4 talks)
    15:20 - 15:50 Mingle in Gather - Meet the industry
    15:50 - 17:15 Industrial Presentations and Award for most industry relevant paper
    17:15 - 17:20 Good Bye - See You Next Year.


Confirmed Keynote speakers



Mattias Ohlsson
  • Mattias Ohlsson
  • Lunds Universitet & Högskolan i Halmstad

  • Mattias Ohlsson is a professor at the department of Astronomy and Theoretical Physics, Lund University. He is also a professor of Information Technology at Halmstad University. His research interests focus on machine learning and its applications in information driven healthcare, ranging from outcome prediction using specific medical registers to analysis of patient trajectories using more comprehensive medical databases. He is further interested in deep generative models and the research around explainable AI.

  • Title: Deep learning in information-driven healthcare: Applications and model development
  • The information-driven care concept is often using electronic health record (EHR) data to build machine learning models for a variety of tasks. Missing data is a reality when working with EHRs, and has to be dealt with prior to machine learning. Here we will look at some deep learning models that can impute missing data, namely Autoencoders and Variational Autoencoders.



Robert Jenssen
  • Robert Jenssen
  • Universitetet i Tromsø

  • Robert Jenssen is Professor and Head of the Machine Learning Group at UiT The Arctic University of Norway: machine-learning.uit.no. He is also an Adjunct Professor at the Norwegian Computing Center in Oslo, Norway. Jenssen received the Dr. Scient (PhD) degree from UiT in 2005. He has had long-term research stays at the University of Florida, at the Technical University of Berlin, and at the Technical University of Denmark. Jenssen's research interests are in the development of novel machine learning methodology, particularly at the intersection of deep learning and kernel machines, with applications in data-driven health technology and in image analysis. Jenssen is a member of the IEEE Technical Committee on Machine Learning for Signal Processing, a member of the Governing Board of IAPR, and an Associate Editor for the journal Pattern Recognition.

  • Title: Industrial and basic deep learning research for computer vision with limited labels
  • This talk takes as a starting point an industrial innovation project for monitoring power lines by deep learning for computer vision. The talk describes briefly a multi-stage pipeline based on traditional supervised object detection and localization for power line monitoring from aerial imagery, developed in collaboration with a company. The need for detecting the power lines themselves is discussed, leading to the development of a novel line segment detector fully utilizing synthetic images, due to the lack of annotations/labeled data for this problem. Further recognizing the importance of learning from limited labeled data, a novel approach to few-shot learning is presented. Finally, going to the extreme case of having no labeled image data available at all, a new deep learning method for multi-view clustering is discussed.



Mattias P. Heinrich
  • Mattias P. Heinrich
  • University of Lübeck
  • Mattias Heinrich is Associate Professor for Medical Image Analysis at the Institute of Medical Informatics at the University of Lübeck. He leads a group on Medical Deep Learning with 10 researchers and enjoys teaching (under)graduate students how to implement solutions to vision problems. Since starting his doctorate in 2009 with Julia Schnabel at the University of Oxford, his passion is medical image registration. He co-organised WBIR 2016 in Las Vegas, the tutorial Learn2Reg at MICCAI 2019 and a multi-task registration challenge at MICCAI 2020. In 2021 he is co-chair of MIDL in Lübeck. He and his team strive to push the boundaries of machine learning and medical image registration and improve its capabilities for clinical applications as well as develop novel theoretical concepts. In 2011 he won the prestigious MICCAI Young Scientist Award for a new descriptor for multimodal registration. His work on large-displacement discrete registration (deeds) led to the 1st place in the EMPIRE10 challenge. To date, he has published 17 MICCAI papers and more than 25 journal articles (many of them in MedIA and TMI with >4000 citations and an h-index of 25). As principal investigator, he has acquired multiple competitive grants (2.0M€ over the last 5 years) for basic research projects mostly involving machine learning and image registration.

  • Title: How to Learn 3D Lung Image Registration? Geometric Learning with Weak Supervision
  • Deep learning has revolutionised image analysis, in particular for segmentation and classification tasks. Medical image registration is still considered challenging for commonly used fully-convolutional deep learning architectures. In this presentation, the advantages of geometric and graphical models within 3D registration will be highlighted for the example of lung registration. Our experiments show surprisingly high performance for aligning 3D point clouds of lungs without any visual features when combining graphical models and geometric learning. Furthermore insights from the ongoing Learn2Reg challenge will be discussed.



Martin Danelljan
  • Martin Danelljan
  • CVL, ETH Zürich

  • Martin Danelljan is a group leader and lecturer at ETH Zürich, Switzerland. He received his Ph.D. degree from Linköping University, Sweden in 2018. His Ph.D. thesis was awarded the biennial Best Nordic Thesis Prize at SCIA 2019. His main research interests are meta and online learning, deep probabilistic models, and conditional generative models. His research includes applications to visual tracking, video object segmentation, dense correspondence estimation, and super-resolution. His research in the field of visual tracking, in particular, has attracted much attention, achieving first rank in the 2014, 2016, and 2017 editions of the Visual Object Tracking (VOT) Challenge and the OpenCV State-of-the-Art Vision Challenge. He received the best paper award at ICPR 2016, the best student paper at BMVC 2019, and an outstanding reviewer award at ECCV 2020. He is also a co-organizer of the VOT, NTIRE, and AIM workshops.

  • Title: Deep Visual Reasoning with Optimization-based Network Modules
  • Deep learning approaches have achieved astonishing performance in numerous vision applications, including image classification, object detection, and semantic segmentation. While these problems are easily treated with standard feed-forward architectures, many computer vision problems require more complex reasoning about the information given during inference. For example, in contrast to semantic segmentation, video object segmentation involves segmenting a set of object instances defined only during inference. An end-to-end learnable approach therefore first needs to extract a powerful representation of the given objects, which can then be used to segment incoming frames. Designing effective end-to-end learnable methods for such tasks have turned out a formidable challenge.
    We tackle this challenge by designing deep network modules that internally optimize an objective. Since key problems in many computer vision tasks are easily formulated as objective functions, optimization-based modules can perform effective and efficient reasoning in such circumstances. By further learning the objective function itself, we obtain a general family of deep network modules, capable of complex non-local reasoning. We will cover their application within a variety of tasks, including visual tracking, video object segmentation, weakly supervised video segmentation, optical flow, and geometric correspondence estimation.



Venue

This year's SSBA/SSDL will be held online through the web platforms Canvas, Zoom and Gather. Information regarding access has been sent out to all registered participants.

If you have any difficulties please contact Niels Christian Overgaard at niels_christian.overgaard@math.lth.se

Organizers

Gold Sponsors

Silver Sponsors

Last Updated Date: 15 MAR 2021