Dreaming more data: Class-dependent distributions over diffeomorphisms for learned data augmentation

Søren Hauberg, Oren Freifeld, Anders Boesen Lindbo Larsen, John W. Fisher, Lars Kai Hansen

Research output: Contribution to conferencePaperpeer-review

73 Scopus citations

Abstract

Data augmentation is a key element in training high-dimensional models. In this approach, one synthesizes new observations by applying pre-specified transformations to the original training data; e.g. new images are formed by rotating old ones. Current augmentation schemes, however, rely on manual specification of the applied transformations, making data augmentation an implicit form of feature engineering. With an eye towards true end-to-end learning, we suggest learning the applied transformations on a per-class basis. Particularly, we align image pairs within each class under the assumption that the spatial transformation between images belongs to a large class of diffeomorphisms. We then learn a class-specific probabilistic generative models of the transformations in a Riemannian submanifold of the Lie group of diffeomorphisms. We demonstrate significant performance improvements in training deep neural nets over manually-specified augmentation schemes. Our code and augmented datasets are available online.

Original languageEnglish
Pages342-350
Number of pages9
StatePublished - 1 Jan 2016
Externally publishedYes
Event19th International Conference on Artificial Intelligence and Statistics, AISTATS 2016 - Cadiz, Spain
Duration: 9 May 201611 May 2016

Conference

Conference19th International Conference on Artificial Intelligence and Statistics, AISTATS 2016
Country/TerritorySpain
CityCadiz
Period9/05/1611/05/16

ASJC Scopus subject areas

  • Artificial Intelligence
  • Statistics and Probability

Fingerprint

Dive into the research topics of 'Dreaming more data: Class-dependent distributions over diffeomorphisms for learned data augmentation'. Together they form a unique fingerprint.

Cite this