Short Seminars by prospective young researches at Computer Science and Engineering - DEIB
Robust Multi Model Fitting in Computer Vision
Luca Magri
Research Assistant - DEIB
Politecnico di Milano, this seminar will be held online
September 23rd, 2020
5.30 pm
Luca Magri
Research Assistant - DEIB
Politecnico di Milano, this seminar will be held online
September 23rd, 2020
5.30 pm
Abstract
Many relevant problems in Computer Vision and Pattern Recognition boil down to performing robust estimation of parametric models (or structures) in data corrupted by noise and outliers. Mainstream algorithms for solving these problems are based on consensus maximization (e.g. RanSaC) and typically fail when data contain multiple structures to be identified. In this seminar, I will present my research activity leading to a new paradigm that builds upon data preferences rather than model consensus, and which develops in a conceptual space where multi-model fitting can be addressed as a clustering problem. This new approach enables a robust, practical and elegant solution to the multi-model fitting problem that automatically estimate the number of structures in the data. The resulting algorithm, T-linkage, has been successfully applied to solve a wide range of problems in Computer Vision and Pattern Recognition, including scan2bim to produce geometric interpretations of the 3D scene or point clouds, motion segmentation in videos and template/object detection. I will conclude my presentation illustrating recent research directions, which relate to the design of anomaly detection algorithms in the preference space, as well as the adoption of more general and flexible data-driven models instead of geometric primitives.
Short Bio
Luca Magri is a postdoc researcher and appointed contract professor at Dipartimento di Elettronica e Informazione e Bioingegneria (DEIB), Politecnico diMilano. His major research interests in clustering techniques and 3D vision intersect pattern recognition and geometric Computer Vision.
The seminar will be held online. Please follow the instructions below:
Meeting number: 121 725 5828
Password: JDxRmeRZ783
https://politecnicomilano.webex.com/
Join by video system
Dial 1217255828@politecnicomilano.webex.com
You can also dial 62.109.219.4 and enter your meeting number.
Join by phone
+44-20-7660-8149 United Kingdom Toll
Access code: 121 725 5828
The seminar will be held online. Please follow the instructions below:
Meeting number: 121 725 5828
Password: JDxRmeRZ783
https://politecnicomilano.webex.com/
Join by video system
Dial 1217255828@politecnicomilano.webex.com
You can also dial 62.109.219.4 and enter your meeting number.
Join by phone
+44-20-7660-8149 United Kingdom Toll
Access code: 121 725 5828