Data Augmentation with Manifold Exploring Geometric Transformations for Increased Performance and Robustness
This work addresses the need for better data augmentation in medical imaging, particularly for limited data scenarios, though it is incremental as it builds on existing manifold exploration methods.
The paper tackles the problem of improving performance and robustness of deep neural networks by proposing a novel data augmentation technique that populates training datasets with images on the border of manifolds between classes, achieving state-of-the-art results in fine-grained skin lesion and breast tumor classification tasks.
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.