Detecting failure modes in image reconstructions with interval neural network uncertainty.

Journal: International journal of computer assisted radiology and surgery
Published Date: Sep 4, 2021

Abstract

PURPOSE: The quantitative detection of failure modes is important for making deep neural networks reliable and usable at scale. We consider three examples for common failure modes in image reconstruction and demonstrate the potential of uncertainty quantification as a fine-grained alarm system.

Authors

  • Luis Oala
    Department of Artificial Intelligence, Fraunhofer HHI, Berlin, Germany. luis.oala@hhi.fraunhofer.de.
  • Cosmas Heiß
    Institut für Mathematik, Technische Universität Berlin, Berlin, Germany.
  • Jan Macdonald
    Institut für Mathematik, Technische Universität Berlin, Berlin, Germany.
  • Maximilian März
    Department of Artificial Intelligence, Fraunhofer HHI, Berlin, Germany.
  • Gitta Kutyniok
    Mathematisches Institut, Ludwig-Maximilians-Universität München, Munich, Germany.
  • Wojciech Samek
    Machine Learning Group, Fraunhofer Heinrich Hertz Institute, Berlin, Germany.

Keywords

External Resources

Popular Topics
Recent Journals