The courses, with both
M.Sc. and Ph.D. students
Researchers and Post-doctoral scholars
Industry professionals with (research) interest in computer vision
... and anyone who wants to discover avant-garde topics
Those with previous contact with the introductory topics of Computer Vision and Machine Learning can expect to enhance their knowledge. Meanwhile, newcomers should see VISUM as an opportunity to get into one of today's most vibrant fields.
VISUM is a safe place for sharing knowledge about science and technology. Want to know more? Check out our
Come learn everything about computer vision from the
Step outside your comfort zone, gather a team, and work together to solve a
How will the challenge work?
the project task will be
participants will be prompted to
teams will be
a
the
mutually exclusive with the mentorship.
VISUM enables you to learn directly from the best! Participants will be able
to select a mentor among a
How will mentorship work?
coaching sessions will be (mostly)
mutually exclusive with the challenge;
mentorship slots are limited.
Is this too tiny to read? Click on the schedule to expand!
How can I apply?
I'm in! How do I register?
If your application has been accepted, we'll contact you with instructions. Keep in mind the following information on fees and deadlines:
Early-bird registrations (until May 6, 2022
Regular registrations (until May 30, 2022
Late registrations (until June 13, 2022
What does the registration include?
all lectures and hands-on sessions;
all workshops;
industry day;
challenge or mentorship;
lunches and coffee breaks;
social programme (on Sunday) and other events;
50% discount on Nilg.ai courses.
How can I get a VISA to attend VISUM?
If you need a visa to travel to Portugal, you need to send us the following information:
your full name;
e-mail address;
address to which you would like the acceptance letter to be sent;
your passport information: number, issue date and place, and expiration date.
Send this info to visum@inesctec.pt with the subject line "Visa letter request". Visas will only be issued after the payment
of registration fees is confirmed.
What else do I need to know?
As a safe place for sharing knowledge about science and technology, all VISUM participants should strictly abide
by the guidelines in our
*All deadlines are 23:59 WEST (UTC+1).
1st Edition
The start of VISUM Summer School
6th Edition
10th Edition
VISUM 2022 is kindly sponsored by
and also supported by
About Ricardo: Ricardo is the Principal Data Scientist at Farfetch. His research interests are Machine Learning and Computer Vision. He has published 30 articles in journals, conferences and book chapters, and has organized five events (being VISUM - VISion Understanding and Machine intelligence - a pivotal summer school in Computer Vision).
Research interests: Conversational Agents, Information Retrieval, Computer Vision.
Most challenging problem he solved: Currently, building a multimodal conversational ai - ifetch-chatbot.github.io.
Selected publications:
> "iFetch: Multimodal Conversational Agents for the Online Fashion Marketplace", R. Sousa et al.
> "Measuring the performance of ordinal classification", J. S. Cardoso and R. Sousa
> "Pose guided attention for multi-label fashion image classification", Q. Ferreira et al.
Recommended papers:
> "A survey on conversational recommender systems", D. Jannach et al.
Mentorship slots: 1 participant.
About Diogo: Diogo Pernes holds a 5-year Master's degree in Electrical and Computers Engineering from the University of Porto (UP) and a PhD in Computer Science at UP. He is an NLP Researcher at Priberam since 2021 and an Invited Teaching Assistant at UP since 2018, teaching courses on algorithms and data structures and probability and statistics. His research has mostly focused on multi-domain learning and domain adaptation using multiple data modalities, including image, text, and data streams.
Research interests: General machine learning/computer vision: domain adaptation, domain generalization, multi-domain learning; Natural Language Processing (NLP): text summarization.
Most challenging problem he solved: The next one is always the most challenging.
Selected publications:
> "SpaMHMM: Sparse Mixture of Hidden Markov Models for Graph Connected Entities", D. Pernes and J. S. Cardoso
> "Tackling unsupervised multi-source domain adaptation with optimism and consistency", D. Pernes and J. S. Cardoso
> "Desire: Deep signer-invariant representations for sign language recognition", P. M. Ferreira et al.
Recommended papers:
> "A theory of learning from different domains", S. Ben-David et al.
> "On Learning Invariant Representation for Domain Adaptation", H. Zhao et al.
> "Neural Text Summarization: A Critical Evaluation", W. Kryściński et al.
Mentorship slots: 1 participant.
About Kelwin: Kelwin Fernandes is the CEO of NILG.AI. He received a PhD in Machine Learning in 2018 from Universidade do Porto and a BSc degree in Computer Engineering in 2012 from Universidad Simón Bolívar. He has worked in AI/ML for the last 12 years, helping dozens of clients to embed data into their daily decisions.
Research interests: Deep Learning, Explainable ML, Self-supervised learning.
Most challenging problem he solved: Soft problems are always the hardest to solve.
Selected publications:
> "Understanding the decisions of CNNs: An in-model approach", I. Rio-Torto et al.
> "Ordinal Image Segmentation using Deep Neural Networks", K. Fernandes and J. S. Cardoso
> "Hypothesis Transfer Learning Based on Structural Model Similarity", K. Fernandes and J. S. Cardoso
Recommended papers:
> "A Survey on Transfer Learning", S. J. Pan and Q. Yang
> "Self-supervised Visual Feature Learning with Deep Neural Networks: A Survey", L. Jing and Y. Tian
Mentorship slots: 2 participants.
About Felix: Felix is a physicist with an engineering PhD and an MBA. He has worked with ML on various data types over the last years, including particle physics data (CERN), hyperspectral remote sensing data from satellites and drones (PhD at KIT), rgb image data, digital health data from smart devices (Roche), and mass spectrometer-based proteomics data (OmicEra Diagnostics). He gave lectures about mathematics, statistics, and ML.
Research interests: Explainable ML, Deep Learning, ML on small data, Self-Organizing Maps, Deployment of ML in production.
Most challenging problem he solved: Merging ML and statistics to fight the curse of dimensionality in proteomics.
Selected publications:
> "Supervised and Semi-Supervised Self-Organizing Maps for Regression and Classification Focusing on Hyperspectral Data", F. M. Riese,
S. Keller, and S. Hinz
> "Soil Texture Classification with 1D Convolutional Neural Networks based on Hyperspectral Data", F. M. Riese and S. Keller
> "Supervised, Semi-supervised, and Unsupervised Learning for Hyperspectral Regression", F. M. Riese and S. Keller
Recommended papers:
> "Interpretable Machine Learning", C. Molnar
Mentorship slots: 1 participant.
About Pedro: Pedro Costa holds a MSc (2015) in Informatics and Computing Engineering at the Faculty of Engineering, University of Porto. Costa started working with INESC TEC in 2014 trying to find adverse drug reactions in biological data using Machine Learning. After a brief experience in the industry, Costa came back to INESC TEC to work on medical image processing using Deep Learning methods, having published several papers in top conferences and medical imaging journals. He then spent three months working on weakly supervised deep learning methods at Carnegie Mellon University (CMU). He was the Head of Research at Abyssal for the past 3 years, having worked to improve the efficiency of Remotely Operated Vehicles' operational efficiency using Machine Learning and Computer Vision. He is currently working at Ocean Infinity, which acquired Abyssal, leading the Artificial Intelligence Research team. The goal of Ocean Infinity is to create the largest fleet of unmanned vessels in the world, and Costa's team is automating all the required processes to make that vision a reality. He is also a PhD student at the University of Porto working on how to reduce the annotation requirements to train Deep Learning models, especially segmentation models.
Research interests: Annotation Efficient Models, Implicit Representations, 3D Computer Vision, Autonomous Underwater Robots.
Most challenging problem he solved: The development of a near real-time SLAM system for underwater environments, working on monocular cameras. The system used very little supervision, and the data used to supervise the model training was very noisy.
Selected publications:
> "Explainable Weakly-Supervised Cell Segmentation by Canonical Shape Learning and Transformation", P. Costa et al.
> "End-to-end adversarial retinal image synthesis", P. Costa et al.
> "A no-reference quality metric for retinal vessel tree segmentation", A. Galdrán et al.
Recommended papers:
> "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis", B. Mildenhall et al.
> "GIRAFFE: Representing Scenes As Compositional Generative Neural Feature Fields", M. Niemeyer and A. Geiger
> "Unsupervised Learning of Depth and Ego-Motion from Video", T. Zhou et al.
Mentorship slots: 3 participants.
About Sailesh: Sailesh Conjeti is currently leading the data science team in Siemens Healthineers. Prior to this, he was the Global Product Manager for AI based image analysis portfolio of applications for radiology. He did his Ph.D. in AI for Medical Imaging from the Technical University of Munich followed by a Post-Doc at the German Center for Neuro-degenerative Diseases. His research interests include deep learning, medical imaging, computational pathology, natural language processing, clinical validation of AI applications and applied machine learning. He has over 50 peer-reviewed publications and has filed over 10 patents in the area of AI/ML, NLP and Medical Imaging. He like to talk about product management, regulatory science, AI algorithm design and validation.
Research interests: Deep learning, medical imaging, computational pathology, natural language processing, clinical validation of AI applications and applied machine learning.
Most challenging problem he solved: Bringing a product from concept/idea to market and going through regulatory clearance - I can talk about the full product lifecycle.
Selected publications:
> "An Artificial Intelligence-Based Chest X-ray Model on Human Nodule Detection Accuracy From a Multicenter Study", F. Homayounieh et al.
> "Fastsurfer-a fast and accurate deep learning based neuroimaging pipeline", L. Henschel et al.
> "Generalizability vs. robustness: adversarial examples for medical imaging", M. Panschali et al.
Recommended papers:
> "Self-supervised Learning from 100 Million Medical Images", F. C. Ghesu et al.
> "Does your dermatology classifier know what it doesn't know? detecting the long-tail of unseen conditions", A. G. Roy et al.
> "How to develop machine learning models for healthcare", P. H. C. Chen et al.
Mentorship slots: 3 participants.
About Mara: Mara is a postdoctoral researcher at IBM Research Zurich and at two universities of applied sciences in Switzerland: ZHAW and HES-SO Valais. She obtained a Ph.D. in Computer Science from the University of Geneva in December 2021. Her research aims at using interpretable deep learning methods to support and facilitate knowledge discovery in bio-medical research. During my Ph.D., she was a visiting student at the Martinos Center, part of the Harvard Medical School in Boston (MA) to focus on the interaction between clinicians and deep learning systems. Coming from a background of IT Engineering, she was awarded the Engineering Department Award for completing the M.Phil. in Machine Learning, Speech and Language Technology at the University of Cambridge (UK) in 2017.
Research interests: Interpretability of Deep Learning, Medical Imaging, Biomedicine, Genomics, Causality, Uncertainty Estimation, and Robustness.
Most challenging problem she solved: Overcoming an unjustified impostor syndrome by learning to celebrate positive feedback just as much as criticisms.
Selected publications:
> "Concept attribution: Explaining CNN decisions to physicians", M. Graziani et al.
> "Learning Interpretable Diagnostic Features of Tumor by Multi-task Adversarial Training of
Convolutional Networks: Improved Generalization", M. Graziani et al.
> "Sharpening Local Interpretable Model-Agnostic Explanations for Histopathology: Improved
Understandability and Reliability", M. Graziani et al.
Recommended papers:
> "Concept whitening for interpretable image recognition", Z. Chen et al.
> "Explaining Classifiers with Causal Concept Effect (CaCE)", Y. Goyal et al.
> "Why Attention is Not Explanation: Surgical Intervention and
Causal Reasoning about Neural Models", C. Grimsley et al.
Mentorship slots: 1 participant.
About Cláudio: Cláudio took a PhD in Computer Science in the University of Leiden (Netherlands). Has more than 11 years of hands-on experience in Data Mining and Machine learning. Worked in different ML fields such as: NLP, Computer Vision, Time-series, Data Mining (Spatio-Temporal mining, Subgroup Discovery), Recommender Systems and Preference Learning.
Research interests: Spatio-temporal Mining, Document Understanding, NLP, Computer Vision, Data Streams.
Most challenging problem he solved: Scaling machine learning approaches.
Selected publications:
> "Variance-based feature importance in neural networks", C. R. de Sá
> "An ensemble of autonomous auto-encoders for human activity recognition", K. Garcia et al.
> "“Want to come play with me?” Outlier subgroup discovery on spatio-temporal interactions", C. Jorge et al.
Recommended papers:
> "Scaling Language Models: Methods, Analysis & Insights from Training Gopher", J. W. Rae et al.
> "LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding", Y. Xu et al.
> "DeepNet: Scaling Transformers to 1,000 Layers", H. Wang et al.
Mentorship slots: 1 participant.
About Dinis: Dinis Moreira holds a MSc (2015) in Bioengineering - Biomedical Engineering Branch at the Faculty of Engineering, University of Porto (FEUP). Moreira started working at Fraunhofer Portugal AICOS in 2015 as a researcher and since then he worked in several different projects within the areas of human motion, mobile/wearable sensors or medical image processing using data mining, classical machine learning and deep learning methods. Since early 2022, he started working for Bosch, within the LIDAR perception team, contributing for the development of AI-related solutions for autonomous driving.
Research interests: Deep Learning, Machine Learning, Explainable AI, Sensors, Time-series, Computer vision.
Most challenging problem he solved: the current/next one because the past ones were already solved.
Selected publications:
> "Human Activity Recognition for Indoor Localization Using Smartphone Inertial Sensors", D. Moreira et al.
> "Automatic focus assessment on dermoscopic images acquired with smartphones", D. Moreira et al.
> "Crowdsourcing-based fingerprinting for indoor location in multi-storey buildings", R. Santos et al.
Recommended papers:
> "Review of deep learning: concepts, CNN architectures, challenges, applications, future directions", L. Alzubaidi et al.
> "Feature Pyramid Networks for Object Detection", T-Y. Lin et al.
> "You Only Look Once: Unified, Real-Time Object Detection", J. Redmon et al.
Mentorship slots: 2 participants.
About Carolina: Carolina Pinto is a Deep Learning Researcher at Bosch since 2020 for interior vehicle sensing and autonomous driving. She completed her M.Sc. in Electrical and Computer Engineering at FEUP in 2020. Her main work conducted at Bosch was focused on using audiovisual information for occupant emotional monitoring, violence detection and activity recognition. Currently, she is focusing her research on the next iteration of perception algorithms for autonomous driving (L4/5), with special attention on temporal and multi-modal fusion networks.
Research interests: Pattern recognition, biometrics, deep learning and computer vision.
Most challenging problem she solved: Developed a functional system prototype for passenger interaction recognition inside a vehicle.
Selected publications:
> "Audiovisual Classification of Group Emotion Valence Using Activity Recognition Networks", J. R. Pinto et al.
> "Impact of visual noise in activity recognition using deep neural networks - an experimental approach", L. Capozzi et al.
> "Streamlining Action Recognition in Autonomous Shared Vehicles with an Audiovisual Cascade Strategy", J. R. Pinto et al.
Recommended papers:
> "Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset", J. Carreira et al.
> "3D-MAN: 3D Multi-frame Attention Network for Object Detection", Z. Yang et al.
> "SalsaNext: Fast, Uncertainty-aware Semantic Segmentation of LiDAR Point Clouds for Autonomous Driving", T. Cortinhal et al.
Mentorship slots: 2 participants.
About João: João Pedro Ribeiro holds a Msc in Electrotechnics and Computing Engineering specialized in Autonomous Systems at Polytechnic of Porto - School of Engineering (ISEP) in 2015. Ribeiro started working in INESC TEC Centre Robotics and Autonomous Systems (CRAS) in 2015 on perception with imaging sensors(EO and IR) in several fields of Autonomous robotics (air, surface and underwater). After that I accepted a new challenge at Bosch Chassis Control as SW developer for LiDAR Perception in Autonomous Driving in topics of Object Detection and Tracking.
Research interests: Robotics, Autonomous Driving, 3D Perception, SW Architecture.
Most challenging problem he solved: The challenge is to be prepared for next problem, because problems can also be seen as opportunities to evolve.
Recommended papers:
> "Simple online and realtime tracking with a deep association metric", N. Wojke et al.
> "Second: Sparsely embedded convolutional detection", Y. Yan et al.
Mentorship slots: 2 participants.
About Henning: Henning Müller studied medical informatics at the University of Heidelberg, Germany, then worked at Daimler-Benz research in Portland, OR, USA. From 1998-2002 he worked on his PhD degree in computer vision at the University of Geneva, Switzerland with a research stay at Monash University, Melbourne, Australia. Since 2002, Henning has been working for the medical informatics service at the University Hospital of Geneva. Since 2007, he has been a full professor at the HES-SO Valais and since 2011 he is responsible for the eHealth unit of the school. Since 2014, he is also professor at the medical faculty of the University of Geneva. In 2015/2016 he was on sabbatical at the Martinos Center, part of Harvard Medical School in Boston, MA, USA to focus on research activities. Henning is coordinator of the ExaMode EU project, was coordinator of the Khresmoi EU project, scientific coordinator of the VISCERAL EU project. Since early 2020 he is also a member of the Swiss National Research Council.
Research interests: machine learning, medical imaging, explainability/interpretability of AI, and medical decision support.
Most challenging problem he solved: problem: Building an integrated histopathology image anaylsis system in the ExaMode project.
Selected publications:
> "A review of content-based image retrieval systems in medicine – clinical benefits and future directions", H. Müller et al.
> "ImageCLEF – Experimental evaluation of visual information retrieval", H. Müller et al.
> "Semi-supervised training of deep convolutional neural networks with heterogeneous data and few local
annotations: an experiment on histopathology image classification", N. Marini et al.
Recommended papers:
> "Content-based image retrieval at the end of the early years", A. W. Smeulders et al.
> "Clinical-grade computational pathology using weakly supervised deep learning on whole slide images", G. Campanella et al.
> "Causability and explainability of artificial intelligence in medicine", A. Holzinger et al.
Mentorship slots: 2 participants.
VISUM is...
a place for sharing knowledge and discussing the future of science and technology;
a place free from discrimination and harassment;
a place for ethical, professional, and appropriate behaviour.
And everyone is expected to work together to keep it that way:
By being part of VISUM Summer School, all participants (including organisers, students, lecturers, mentors, industry representatives, sponsors, partners, and everyone else attending or contributing to VISUM, or in any way related to this summer school) agree to follow this code of conduct during the entirety of VISUM's online and on-site activities and on all activities on official communication channels, including on social media.
Harassment, discrimination, and bullying will not be tolerated. This includes all kinds of offensive comments related to age, race, ethnicity, religion, creed, colour, gender, sexual orientation, medical condition, physical or intellectual disabilities, pregnancy, national origin, or ancestry.
Inappropriate or unprofessional behaviours have no place at VISUM. This includes sexual harassment, stalking, harassing photography or recording, inappropriate physical contact or attention, public vulgar exchanges, derogatory name-calling, and diminutive characterisations. This also includes the use, display, or unsolicited sharing of images, activities, or other materials of sexual, racial, or otherwise offensive nature.
VISUM is a place for orderly, polite, and respectful behaviour. Disorderly conduct such as fighting, insult, coercion, theft, or damage to property will not be tolerated. Similarly, any action aimed at disturbing the correct operation of VISUM activities will also not be tolerated.
VISUM is not a place for political discussion. All participants should refrain from expressing or discussing political perspectives or viewpoints not directly related to the VISUM topics of computer vision and machine learning. Additionally, any political viewpoints or beliefs held by participants are their own and are not in any way endorsed by VISUM Summer School.
Privacy is of paramount importance. All participants should be careful not to disclose others' private or intimate information, including images or recordings, without prior authorisation. This includes the sharing of photographs, videos, or audio excerpts captured during VISUM's activities.
We reserve the right to take action against offenders:
VISUM organisers may take action against participants breaking any rule in this code of conduct as deemed appropriate. This includes formal or informal warnings; expulsion from the summer school with no refund; barring from participation in future editions of VISUM; reporting the incident to the offender's affiliation institution or funding agencies; or reporting the incident to local authorities or law enforcement.
If action is taken, an appeals process will be made available.
We expect every participant to assist us in keeping VISUM safe:
Organisers encourage all participants to immediately report any incidents of discrimination, harassment, unprofessional conduct, retaliation, or any other behaviour breaking the rules in this code of conduct.
Complaints and incident reports may be delivered on-site to any member of the VISUM organising team, appropriately identified on our website and through worn nametags, or by email to visum@inesctec.pt.
All complaints or incident reports submitted to the organisers will be immediately and thoroughly investigated. There will be no retaliation against any participant who raises a complaint or submits an incident report in good faith or who honestly assists in any investigation.
