D2
Computer Vision and Machine Learning

Moritz Böhle (PhD Student)

MSc Moritz Daniel Böhle

Adresse
Max-Planck-Institut für Informatik
Saarland Informatics Campus
Campus E1 4
66123 Saarbrücken
Standort
E1 4 - 622
Telefon
+49 681 9325 2122
Fax
+49 681 9325 2099

Personal Information

Publications

Böhle, M. D., Fritz, M., & Schiele, B. (2021a). Convolutional Dynamic Alignment Networks for Interpretable Classifications. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2021). Nashville, TN, USA (Virtual): IEEE. doi:10.1109/CVPR46437.2021.00990
[DOI]
[PuRe]
[BibTeX]
[pre-print version]
Links
[DOI]
[PuRe]
[BibTeX]
[pre-print version]
Export
BibTeX
@inproceedings{Boehle_CVPR21, TITLE = {Convolutional Dynamic Alignment Networks for Interpretable Classifications}, AUTHOR = {B{\"o}hle, Moritz Daniel and Fritz, Mario and Schiele, Bernt}, LANGUAGE = {eng}, ISBN = {978-1-6654-4509-2}, DOI = {10.1109/CVPR46437.2021.00990}, PUBLISHER = {IEEE}, YEAR = {2021}, MARGINALMARK = {$\bullet$}, BOOKTITLE = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2021)}, PAGES = {10029--10038}, ADDRESS = {Nashville, TN, USA (Virtual)}, }
Endnote
%0 Conference Proceedings %A Böhle, Moritz Daniel %A Fritz, Mario %A Schiele, Bernt %+ Computer Vision and Machine Learning, MPI for Informatics, Max Planck Society External Organizations Computer Vision and Machine Learning, MPI for Informatics, Max Planck Society %T Convolutional Dynamic Alignment Networks for Interpretable Classifications : %G eng %U http://hdl.handle.net/21.11116/0000-0008-1863-E %R 10.1109/CVPR46437.2021.00990 %D 2021 %B 34th IEEE Conference on Computer Vision and Pattern Recognition %Z date of event: 2021-06-19 - 2021-06-25 %C Nashville, TN, USA (Virtual) %B IEEE/CVF Conference on Computer Vision and Pattern Recognition %P 10029 - 10038 %I IEEE %@ 978-1-6654-4509-2
Böhle, M. D., Fritz, M., & Schiele, B. (2021b). Optimising for Interpretability: Convolutional Dynamic Alignment Networks. Retrieved from https://arxiv.org/abs/2109.13004
(arXiv: 2109.13004)
[arXiv]
[PuRe]
[BibTeX]
[pre-print version]
Abstract
We introduce a new family of neural network models called Convolutional Dynamic Alignment Networks (CoDA Nets), which are performant classifiers with a high degree of inherent interpretability. Their core building blocks are Dynamic Alignment Units (DAUs), which are optimised to transform their inputs with dynamically computed weight vectors that align with task-relevant patterns. As a result, CoDA Nets model the classification prediction through a series of input-dependent linear transformations, allowing for linear decomposition of the output into individual input contributions. Given the alignment of the DAUs, the resulting contribution maps align with discriminative input patterns. These model-inherent decompositions are of high visual quality and outperform existing attribution methods under quantitative metrics. Further, CoDA Nets constitute performant classifiers, achieving on par results to ResNet and VGG models on e.g. CIFAR-10 and TinyImagenet. Lastly, CoDA Nets can be combined with conventional neural network models to yield powerful classifiers that more easily scale to complex datasets such as Imagenet whilst exhibiting an increased interpretable depth, i.e., the output can be explained well in terms of contributions from intermediate layers within the network.
Links
[PuRe]
[arXiv]
[BibTeX]
[pre-print version]
Export
BibTeX
@online{Boehle2109.13004, TITLE = {Optimising for Interpretability: Convolutional Dynamic Alignment Networks}, AUTHOR = {B{\"o}hle, Moritz Daniel and Fritz, Mario and Schiele, Bernt}, LANGUAGE = {eng}, URL = {https://arxiv.org/abs/2109.13004}, EPRINT = {2109.13004}, EPRINTTYPE = {arXiv}, YEAR = {2021}, MARGINALMARK = {$\bullet$}, ABSTRACT = {We introduce a new family of neural network models called Convolutional Dynamic Alignment Networks (CoDA Nets), which are performant classifiers with a high degree of inherent interpretability. Their core building blocks are Dynamic Alignment Units (DAUs), which are optimised to transform their inputs with dynamically computed weight vectors that align with task-relevant patterns. As a result, CoDA Nets model the classification prediction through a series of input-dependent linear transformations, allowing for linear decomposition of the output into individual input contributions. Given the alignment of the DAUs, the resulting contribution maps align with discriminative input patterns. These model-inherent decompositions are of high visual quality and outperform existing attribution methods under quantitative metrics. Further, CoDA Nets constitute performant classifiers, achieving on par results to ResNet and VGG models on e.g. CIFAR-10 and TinyImagenet. Lastly, CoDA Nets can be combined with conventional neural network models to yield powerful classifiers that more easily scale to complex datasets such as Imagenet whilst exhibiting an increased interpretable depth, i.e., the output can be explained well in terms of contributions from intermediate layers within the network.}, }
Endnote
%0 Report %A Böhle, Moritz Daniel %A Fritz, Mario %A Schiele, Bernt %+ Computer Vision and Machine Learning, MPI for Informatics, Max Planck Society External Organizations Computer Vision and Machine Learning, MPI for Informatics, Max Planck Society %T Optimising for Interpretability: Convolutional Dynamic Alignment Networks : %G eng %U http://hdl.handle.net/21.11116/0000-0009-8113-F %U https://arxiv.org/abs/2109.13004 %D 2021 %X We introduce a new family of neural network models called Convolutional Dynamic Alignment Networks (CoDA Nets), which are performant classifiers with a high degree of inherent interpretability. Their core building blocks are Dynamic Alignment Units (DAUs), which are optimised to transform their inputs with dynamically computed weight vectors that align with task-relevant patterns. As a result, CoDA Nets model the classification prediction through a series of input-dependent linear transformations, allowing for linear decomposition of the output into individual input contributions. Given the alignment of the DAUs, the resulting contribution maps align with discriminative input patterns. These model-inherent decompositions are of high visual quality and outperform existing attribution methods under quantitative metrics. Further, CoDA Nets constitute performant classifiers, achieving on par results to ResNet and VGG models on e.g. CIFAR-10 and TinyImagenet. Lastly, CoDA Nets can be combined with conventional neural network models to yield powerful classifiers that more easily scale to complex datasets such as Imagenet whilst exhibiting an increased interpretable depth, i.e., the output can be explained well in terms of contributions from intermediate layers within the network. %K Statistics, Machine Learning, stat.ML,Computer Science, Computer Vision and Pattern Recognition, cs.CV,Computer Science, Learning, cs.LG