ir. W.P. (Willem) Sanberg - Expertises

Sanberg, ir. W.P.
Adres :
Technische Universiteit Eindhoven
Postbus 513
5600 MB EINDHOVEN
Faculteit :
Faculteit Electrical Engineering
Afdeling :
Video Coding & Architectures
Functiecategorie :
Postdoc (PD)
Functie :
Postdoc
Intern adres :
FLX 5.092
Email :
wsanberg@tue.nl

Expertises

  • Machine learning
  • Computer-ondersteunde diagnose

Biografie

Academic background

Willem Sanberg received both his BSc. (2011) and his MSc.-degree (2013) in Electrical Engineering and a certificate in Technical Management (2011) from Eindhoven University of Technology (TU/e) in the Netherlands. Furthermore, he was active in committees and boards of a student association for several years, a member of University Racing Eindhoven for two years and a student-assistant at the electronics lab of Industrial Design for two years as well. During his MSc.-education, he was an intern at Philips Research for several months, developing an AdaBoost detector for facial hair detection in 3D face meshes. He graduated on the topic of Graph-based Multi-Modal segmentation at the VCA research group. After his graduation, he stayed with VCA as a PhD candidate. His current research is aimed at improving sensing capabilities of mobile platforms using multiple sensors that capture different modalities. On such platform, data should be captured and processed in real-time, so efficient data fusion methods –both in time and over modalities- will need to be developed.

Research profile

Willem's current research addresses improving the perception of Advanced Driver Assistance Systems (ADAS) for intelligent vehicles. His main sensors of interest are stereo vision cameras, providing rich and dense data on the appearance and geometry of traffic scenes.

He works on algorithms that leverage those two data stream individually and combined to improve free-space segmentation. The core is a self-supervised method that adapts itself to the changing scenery while the vehicle is driving, which offers a trade-off between efficient processing and flexibility to new environments. It is a combination of classical computer vision and trendy convolutional neural nets.

Additionally, his recent works looks into the dual problem of free-space segmentation: object detection. It encompasses a collision warning system based on stereo vision.