Latent Discriminant deterministic Uncertainty
Predictive uncertainty estimation is essential for deploying Deep Neural Networks in real-world autonomous systems. However, most successful approaches are computationally intensive. In this seminar Dr. Gianni Franchi will attempt to address these challenges in the context of autonomous driving perception tasks. Recently proposed Deterministic Uncertainty Methods (DUM) can only partially meet such requirements as their scalability to complex computer vision tasks is not obvious. In this work, a scalable and effective DUM for high-resolution semantic segmentation that relaxes the Lipschitz constraint has been advanced, typically hindering practicality of such architectures. A discriminant latent space by leveraging a distinction maximization layer over an arbitrarily-sized set of trainable prototypes has been learnt. The approach achieves competitive results over Deep Ensembles, the state-of-the-art for uncertainty prediction, on image classification, segmentation, and monocular depth estimation tasks.
About the Speaker
Dr. Gianni Franchi received an MSc degree in engineering science from Ecole Centrale Marseille in 2013, and a Ph.D. degree in applied mathematics (2016) from PSL University (Mines ParisTech). Then he did a Postdoc at Siegen University from October 2016 to December 2017. Then he did a Postdoc a Paris Saclay University.
He is currently pursuing an Assistant Professor at ENSTA Paris. His topics of interest include uncertainty quantification, Semi-Supervised Learning, Domain adaptation, computer vision, machine learning, and statistical learning.