Many real-world vision problems suffer from inherent ambiguities. In clinical
applications for example, it might not be clear from a CT scan alone which
particular region is cancer tissue. Therefore a group of graders typically
produces a set of diverse but plausible segmentations. We consider the task of
learning a distribution over segmentations given an input. To this end we
propose a generative segmentation model based on a combination of a U-Net with
a conditional variational autoencoder that is capable of efficiently producing
an unlimited number of plausible hypotheses. We show on a lung abnormalities
segmentation task and on a Cityscapes segmentation task that our model
reproduces the possible segmentation variants as well as the frequencies with
which they occur, doing so significantly better than published approaches.
These models could have a high impact in real-world applications, such as being
used as clinical decision-making algorithms accounting for multiple plausible
semantic segmentation hypotheses to provide possible diagnoses and recommend
further actions to resolve the present ambiguities.
cite arxiv:1806.05034Comment: Last update: change to new NeurIPS style. 11 pages for the main paper, 28 pages including appendix. 5 figures in the main paper, 18 figures in total, Advances in Neural Information Processing Systems (NeurIPS), 2018
%0 Generic
%1 kohl2018probabilistic
%A Kohl, Simon A. A.
%A Romera-Paredes, Bernardino
%A Meyer, Clemens
%A De Fauw, Jeffrey
%A Ledsam, Joseph R.
%A Maier-Hein, Klaus H.
%A Eslami, S. M. Ali
%A Rezende, Danilo Jimenez
%A Ronneberger, Olaf
%D 2018
%K cnn
%T A Probabilistic U-Net for Segmentation of Ambiguous Images
%U http://arxiv.org/abs/1806.05034
%X Many real-world vision problems suffer from inherent ambiguities. In clinical
applications for example, it might not be clear from a CT scan alone which
particular region is cancer tissue. Therefore a group of graders typically
produces a set of diverse but plausible segmentations. We consider the task of
learning a distribution over segmentations given an input. To this end we
propose a generative segmentation model based on a combination of a U-Net with
a conditional variational autoencoder that is capable of efficiently producing
an unlimited number of plausible hypotheses. We show on a lung abnormalities
segmentation task and on a Cityscapes segmentation task that our model
reproduces the possible segmentation variants as well as the frequencies with
which they occur, doing so significantly better than published approaches.
These models could have a high impact in real-world applications, such as being
used as clinical decision-making algorithms accounting for multiple plausible
semantic segmentation hypotheses to provide possible diagnoses and recommend
further actions to resolve the present ambiguities.
@misc{kohl2018probabilistic,
abstract = {Many real-world vision problems suffer from inherent ambiguities. In clinical
applications for example, it might not be clear from a CT scan alone which
particular region is cancer tissue. Therefore a group of graders typically
produces a set of diverse but plausible segmentations. We consider the task of
learning a distribution over segmentations given an input. To this end we
propose a generative segmentation model based on a combination of a U-Net with
a conditional variational autoencoder that is capable of efficiently producing
an unlimited number of plausible hypotheses. We show on a lung abnormalities
segmentation task and on a Cityscapes segmentation task that our model
reproduces the possible segmentation variants as well as the frequencies with
which they occur, doing so significantly better than published approaches.
These models could have a high impact in real-world applications, such as being
used as clinical decision-making algorithms accounting for multiple plausible
semantic segmentation hypotheses to provide possible diagnoses and recommend
further actions to resolve the present ambiguities.},
added-at = {2018-12-15T04:40:48.000+0100},
author = {Kohl, Simon A. A. and Romera-Paredes, Bernardino and Meyer, Clemens and De Fauw, Jeffrey and Ledsam, Joseph R. and Maier-Hein, Klaus H. and Eslami, S. M. Ali and Rezende, Danilo Jimenez and Ronneberger, Olaf},
biburl = {https://www.bibsonomy.org/bibtex/2300edcc964011b7550ff1bbf7053571a/zezhou},
description = {1806.05034.pdf},
interhash = {3c9e109565b551e82f2137010d130c3e},
intrahash = {300edcc964011b7550ff1bbf7053571a},
keywords = {cnn},
note = {cite arxiv:1806.05034Comment: Last update: change to new NeurIPS style. 11 pages for the main paper, 28 pages including appendix. 5 figures in the main paper, 18 figures in total, Advances in Neural Information Processing Systems (NeurIPS), 2018},
timestamp = {2018-12-15T04:40:48.000+0100},
title = {A Probabilistic U-Net for Segmentation of Ambiguous Images},
url = {http://arxiv.org/abs/1806.05034},
year = 2018
}