Data Augmentation Generative Adversarial Networks

By Antreas Antoniou, Amos J. Storkey, Harrison Edwards. 2017

1,180 citations recorded by the source corpus.

Effective training of neural networks requires much data. In the low-data regime, parameters are underdetermined, and learnt networks generalise poorly. Data Augmentation alleviates this by using existing data more effectively. However standard data augmentation produces only limited plausible alternative data. Given there is potential to generate a much broader set of augmentations, we design and train a generative model to do data augmentation. The model, based on image conditional Generative Adversarial Networks, takes data from a source domain and learns to take any data item and generalise it to generate other within-class data items. As this generative process does not depend on the classes themselves, it can be applied to novel unseen classes of data. We show that a Data Augmentation Generative Adversarial Network (DAGAN) augments standard vanilla classifiers well. We also show a DAGAN can enhance few-shot learning systems such as Matching Networks. We demonstrate these approaches on Omniglot, on EMNIST having learnt the DAGAN on Omniglot, and VGG-Face data. In our experiments we can see over 13% increase in accuracy in the low-data regime experiments in Omniglot (from 69% to 82%), EMNIST (73.9% to 76%) and VGG-Face (4.5% to 12%); in Matching Networks for Omniglot we observe an increase of 0.5% (from 96.9% to 97.4%) and an increase of 1.8% in EMNIST (from 59.5% to 61.3%).

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Generative Models

Citations in the corpus

Break-It-Fix-It: Unsupervised Learning for Program Repair (2021, methodology): “A common approach is to increase the source side data, for instance by adding modifications or sampling from generative models (Hannun et al., 2014; Jia & Liang, 2016; Krizhevsky et al., 2017; Antoniou et al., 2017; Yasunaga et al., 2018; Yu et al., 2018a; Lee et al., 2019; Berthelot et al., 2019; X”
Randaugment: Practical automated data augmentation with a reduced search space (2019, methodology): “GANs have also been used to generate training data directly [31, 27, 50, 1, 37], however this approach does not seem to be as beneficial as learning sequences of data augmentation operations that are pre-defined [33].”
How to train your MAML (2018, background): “state of the art results in a variety of settings (Wang et al., 2016; Ba et al., 2016; Zoph & Le, 2016; Andrychowicz et al., 2016; Vinyals et al., 2016; Li & Malik, 2016; Li et al., 2017; Antoniou et al., 2017; Brock et al., 2017; Munkhdalai & Yu, 2017).”
A Closer Look at Few-shot Classification (2019, methodology): “(2018), and hallucination based methods Antoniou et al. (2018); Hariharan & Girshick (2017); Wang et al.”
Unsupervised Adversarial Invariance (2018, methodology): “A popular approach to learn models robust to such expected yet unobserved or infrequently seen (during training) variations is data augmentation through synthetic generation using methods ranging from simple operations [10] like rotation and translation to Generative Adversarial Networks [1, 8] for ”
Adversarial AutoAugment (2019, methodology): “…et al., 2014), which has recently attracted lots of research attention due to its fascinating performance, and also been used to enlarge datasets through directly synthesizing new images (Tran et al., 2017; Perez & Wang, 2017; Antoniou et al., 2017; Gurumurthy et al., 2017; Frid-Adar et al., 2018).”
Faster AutoAugment: Learning Augmentation Strategies using Backpropagation (2019, methodology): “gle in rotate. On the other hand, some operations, such as invert, have no magnitude parameter. For simplicity, we use the following expressions as if every operation had its magnitude parameter O(2[0;1]). Each operation is applied with probability of p O(2[0;1]). Therefore, each image Xis augmente”
Automatically Learning Data Augmentation Policies for Dialogue Tasks (2019, background): “…have also adopted Generative Adversarial Networks (Goodfellow et al., 2014) to either directly generate augmented data (Tran et al., 2017; Sixt et al., 2018; Antoniou et al., 2017; Zhu et al., 2017; Mun et al., 2017; Wang and Perez, 2017), or generate augmentation strategies (Ratner et al., 2017).”
Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules (2019, background): “Some additional approaches to generate augmented data include image combining (Inoue, 2018; Okafor et al., 2018), elastic distortions (Wong et al., 2016), and generative adversarial networks (Antoniou et al., 2017).”
GAN Augmentation: Augmenting Training Data using Generative Adversarial Networks (2018, methodology): “In [1], the authors train a conditional GAN on unlabelled data to generate alternative versions of a given real image, and in [23], the authors use a similar GAN to impose emotions on neutral faces to expand underrepresented classes.”
Continual Learning through Retrieval and Imagination (2022, background): “However, for some complex datasets with large intra-class distances, such as CI-FAR (Krizhevsky et al. 2009) and ImageNet (Deng et al. 2009), GAN will collapse on them because it has to learn the diversity of the huge differences (Antoniou et al. 2017).”
Defining Benchmarks for Continual Few-Shot Learning (2020, methodology): “An example of this approach has been recently presented by Antoniou et al. (2017).”
Meta-Learning in Neural Networks: A Survey (2020, methodology): “based data augmentation methods [147] can be used in inner-level learning and optimized in outer-level learning.”
Learning Data Augmentation Strategies for Object Detection (2019, methodology): “For image classification models, data can be augmented either by learning a generator that can create data from scratch [33, 31, 54, 2, 43], or by learning a set of transformations as applied to already existing training set samples [5, 37].”
Bias Correction of Learned Generative Models using Likelihood-Free Importance Weighting (2019, methodology): “While DAGAN was motivated by and evaluated for the task of meta-learning, it can also be applied for multi-class classiﬁcation scenarios, which is the setting we consider here.”
Interventional Few-Shot Learning (2020, methodology): “These methods learn to generate additional support samples with image deformation [12, 74] or generative models [2, 75].”
Learning More with Less: Conditional PGGAN-based Data Augmentation for Brain Metastases Detection Using Highly-Rough Annotation on MR Images (2019, background): “…[19] suffering from the injected noise and imperfect reconstruction due to a single objective function [22]; thus, as a DA technique, most researchers chose GANs for facilitating classification [1, 21], object detection [12, 24], and segmentation [38] to tackle the lack of training data.”