Timezone: »

 
Poster
Correlated Uncertainty for Learning Dense Correspondences from Noisy Labels
Natalia Neverova · David Novotny · Andrea Vedaldi

Wed Dec 11 05:00 PM -- 07:00 PM (PST) @ East Exhibition Hall B + C #82
in Applications -- Computer Vision »

Many machine learning methods depend on human supervision to achieve optimal performance. However, in tasks such as DensePose, where the goal is to establish dense visual correspondences between images, the quality of manual annotations is intrinsically limited. We address this issue by augmenting neural network predictors with the ability to output a distribution over labels, thus explicitly and introspectively capturing the aleatoric uncertainty in the annotations. Compared to previous works, we show that correlated error fields arise naturally in applications such as DensePose and these fields can be modeled by deep networks, leading to a better understanding of the annotation errors. We show that these models, by understanding uncertainty better, can solve the original DensePose task more accurately, thus setting the new state-of-the-art accuracy in this benchmark. Finally, we demonstrate the utility of the uncertainty estimates in fusing the predictions of produced by multiple models, resulting in a better and more principled approach to model ensembling which can further improve accuracy.

Keywords: Computer Vision · Applications · Algorithms · Uncertainty Estimation
[ Paper]  [ Slides] 

Author Information

Natalia Neverova (Facebook AI Research)
David Novotny (Facebook AI Research)
Andrea Vedaldi (University of Oxford / Facebook AI Research)

More from the Same Authors