Abstract In numerous real world applications, no matter how much energy is devoted to build real and/or synthetic training datasets, there remains a large distribution gap between these data and those met at run-time. This gap results in severe, possibly catastrophic, performance loss. This problem is especially acute for automated and autonomous driving systems, where generalizing well to diverse testing environments remains a major challenge. One promising tool to mitigate this issue it unsupervised domain adaptation (UDA), which assumes that un-annotated data from the “test domain” are available at training time, along with the annotated data from the “source domain”. We will discuss different ways to approach UDA, with application to semantic segmentation and object detection in urban scenes. We will introduce a new approach, called AdvEnt, that relies on combining adversarial training with minimization of decision entropy (seen as a proxy for uncertainty).
[Event]
[Slides]

Attacking inverse problems with deep learning

March 21, 2019

Research, Artificial intelligence and physics, Saclay, France

Abstract [TBC]
[Event]
[Slides]

Patrick Pérez

Talks and presentations

Sustainable supervision with application to autonomous driving

Sustainable supervision with application to autonomous driving

Unsupervised domain adaptation with application to urban scene analysis

Deep Learning for inverse problems

ADVENT: Adversarial Entropy Minimization for domain adaptation in semantic segmentation

Unsupervised domain adaptation with application to urban scene analysis

Attacking inverse problems with deep learning

Movies & Cars: When deep learning reinvents two secular industries

Introduction to Deep learning

Deep learning for the movie industry

Introduction to deep learning

Deep learning for the movie industry

Graph signals and neural nets

Graph signals and neural nets

Deep learning for the movie industry

From processing to learning on graphs

From processing to learning on graphs

AI and Industry