Manifold Exploring Data Augmentation with Geometric Transformations for Increased Performance and Robustness
Paschali M, Simson W, Roy AG, Goebl R, Wachinger C, Navab N (2019)
Publication Type: Conference contribution
Publication year: 2019
Journal
Publisher: Springer Verlag
Book Volume: 11492 LNCS
Pages Range: 517-529
Conference Proceedings Title: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Event location: Hong Kong, CHN
ISBN: 9783030203504
DOI: 10.1007/978-3-030-20351-1_40
Abstract
In this paper we propose a novel augmentation technique that improves not only the performance of deep neural networks on clean test data, but also significantly increases their robustness to random transformations, both affine and projective. Inspired by ManiFool, the augmentation is performed by a line-search manifold-exploration method that learns affine geometric transformations that lead to the misclassification on an image, while ensuring that it remains on the same manifold as the training data. This augmentation method populates any training dataset with images that lie on the border of the manifolds between two-classes and maximizes the variance the network is exposed to during training. Our method was thoroughly evaluated on the challenging tasks of fine-grained skin lesion classification from limited data, and breast tumor classification of mammograms. Compared with traditional augmentation methods, and with images synthesized by Generative Adversarial Networks our method not only achieves state-of-the-art performance but also significantly improves the network’s robustness.
Involved external institutions
Germany (DE)How to cite
APA:
Paschali, M., Simson, W., Roy, A.G., Goebl, R., Wachinger, C., & Navab, N. (2019). Manifold Exploring Data Augmentation with Geometric Transformations for Increased Performance and Robustness. In Albert C.S. Chung, Siqi Bao, James C. Gee, Paul A. Yushkevich (Eds.), Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (pp. 517-529). Hong Kong, CHN: Springer Verlag.
MLA:
Paschali, Magdalini, et al. "Manifold Exploring Data Augmentation with Geometric Transformations for Increased Performance and Robustness." Proceedings of the 26th International Conference on Information Processing in Medical Imaging, IPMI 2019, Hong Kong, CHN Ed. Albert C.S. Chung, Siqi Bao, James C. Gee, Paul A. Yushkevich, Springer Verlag, 2019. 517-529.
BibTeX: Download