Dr. Wouter Van Gansbeke

I will be a tenure-track faculty member at INSAIT. 

Contact: [email protected]
Google Scholar: link

My main research interests lie in the area of computer vision. In particular, I currently work on the following domains and their intersections:

Understanding Visual Scenes: Modeling geometry, discovering objects and patterns in complex scenes;
Self-Supervised Learning: Learn useful representations or groups without annotations by leveraging visual similarities;
Generative Learning: Capture the underlying data distribution of images using generative principles;
Multi-Modal Learning:  Learn from different modalities, or streams of information, since our world is inherently multi-modal, e.g., leverage vision and text.

I strongly believe that these fields are important to enable machine intelligence and large-scale autonomy as already evidenced in applications such as autonomous driving and augmented reality (AR / VR). If you’re interested in any of the mentioned topics, don’t hesitate to reach out.


Academic degrees:

PhD degree in Electrical Engineering from KU Leuven, Belgium at the Center of Processing Speech and Images:
Supervisor: Prof. Luc Van Gool
Thesis Title: Exploring Unsupervised Learning for Computer Vision Tasks with Neural Networks
Project: Toyota R&D project – TRACE (Toyota Research for Automated Cars in Europe), coordinated by Dr. Marc Proesmans

Master’s and Bachelor’s degree in Engineering Science from KU Leuven, Belgium:
Master’s Thesis: Real-time Scene Understanding for Autonomous Driving, supervised by Prof. Luc Van Gool

Recognition – Research Community:

Outstanding Reviewer for ECCV 2022 and CVPR 2023.
Our ECCV paper received attention from the research community, with multiple YouTube videos and blog posts discussing its findings.
Co-organized the workshop ”Deep Multi-Task Learning in Computer Vision (DeepMTL)” at ICCV 2021.

Before, I interned at Meta (Facebook AI Research) in California.