When we are faced with challenging image classification tasks, we often
explain our reasoning by dissecting the image, and pointing out prototypical
aspects of one class or another. The mounting evidence for each of the classes
helps us make our final decision. In this work, we introduce a deep network
architecture that reasons in a similar way: the network dissects the image by
finding prototypical parts, and combines evidence from the prototypes to make a
final classification. The algorithm thus reasons in a way that is qualitatively
similar to the way ornithologists, physicians, geologists, architects, and
others would explain to people on how to solve challenging image classification
tasks. The network uses only image-level labels for training, meaning that
there are no labels for parts of images. We demonstrate the method on the
CIFAR-10 dataset and 10 classes from the CUB-200-2011 dataset.
Description
[1806.10574] This looks like that: deep learning for interpretable image recognition
%0 Generic
%1 chen2018looks
%A Chen, Chaofan
%A Li, Oscar
%A Barnett, Alina
%A Su, Jonathan
%A Rudin, Cynthia
%D 2018
%K 2018 arxiv computer-vision deep-learning paper
%T This looks like that: deep learning for interpretable image recognition
%U http://arxiv.org/abs/1806.10574
%X When we are faced with challenging image classification tasks, we often
explain our reasoning by dissecting the image, and pointing out prototypical
aspects of one class or another. The mounting evidence for each of the classes
helps us make our final decision. In this work, we introduce a deep network
architecture that reasons in a similar way: the network dissects the image by
finding prototypical parts, and combines evidence from the prototypes to make a
final classification. The algorithm thus reasons in a way that is qualitatively
similar to the way ornithologists, physicians, geologists, architects, and
others would explain to people on how to solve challenging image classification
tasks. The network uses only image-level labels for training, meaning that
there are no labels for parts of images. We demonstrate the method on the
CIFAR-10 dataset and 10 classes from the CUB-200-2011 dataset.
@misc{chen2018looks,
abstract = {When we are faced with challenging image classification tasks, we often
explain our reasoning by dissecting the image, and pointing out prototypical
aspects of one class or another. The mounting evidence for each of the classes
helps us make our final decision. In this work, we introduce a deep network
architecture that reasons in a similar way: the network dissects the image by
finding prototypical parts, and combines evidence from the prototypes to make a
final classification. The algorithm thus reasons in a way that is qualitatively
similar to the way ornithologists, physicians, geologists, architects, and
others would explain to people on how to solve challenging image classification
tasks. The network uses only image-level labels for training, meaning that
there are no labels for parts of images. We demonstrate the method on the
CIFAR-10 dataset and 10 classes from the CUB-200-2011 dataset.},
added-at = {2018-09-19T17:35:04.000+0200},
author = {Chen, Chaofan and Li, Oscar and Barnett, Alina and Su, Jonathan and Rudin, Cynthia},
biburl = {https://www.bibsonomy.org/bibtex/27d9186d6e10ab96dc2346af60b9193c9/analyst},
description = {[1806.10574] This looks like that: deep learning for interpretable image recognition},
interhash = {308107426c54585a5f45788ced4d791e},
intrahash = {7d9186d6e10ab96dc2346af60b9193c9},
keywords = {2018 arxiv computer-vision deep-learning paper},
note = {cite arxiv:1806.10574Comment: 14 pages, including the supplementary material},
timestamp = {2018-09-19T17:35:04.000+0200},
title = {This looks like that: deep learning for interpretable image recognition},
url = {http://arxiv.org/abs/1806.10574},
year = 2018
}