BAGAN: Data Augmentation with Balancing GAN

By Giovanni Mariani, Florian Scheidegger, Roxana Istrate, C. Bekas, A. Malossi. 2018

357 citations recorded by the source corpus.

Image classification datasets are often imbalanced, characteristic that negatively affects the accuracy of deep-learning classifiers. In this work we propose balancing GAN (BAGAN) as an augmentation tool to restore balance in imbalanced datasets. This is challenging because the few minority-class images may not be enough to train a GAN. We overcome this issue by including during the adversarial training all available images of majority and minority classes. The generative model learns useful features from majority classes and uses these to generate images for minority classes. We apply class conditioning in the latent space to drive the generation process towards a target class. The generator in the GAN is initialized with the encoder module of an autoencoder that enables us to learn an accurate class-conditioning in the latent space. We compare the proposed methodology with state-of-the-art GANs and demonstrate that BAGAN generates images of superior quality when trained with an imbalanced dataset.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Citations in the corpus

Data Augmentation Using GANs (2019, methodology): “Regarding the two applications addressed in this work (balancing data and generating synthetic data sets), there is some previous work in the literature using GANs, such as Mariani et al. (2018), Springenberg (2015) and Frid-Adar et al. (2018)).”
Learning More with Less: Conditional PGGAN-based Data Augmentation for Brain Metastases Detection Using Highly-Rough Annotation on MR Images (2019, background): “GANs can typically generate more realistic images than other common deep generative models, such as variational autoencoders [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 facilitat”
Unsupervised Attention-guided Image to Image Translation (2018, background): “Applications include image colorization [6], image super-resolution [7, 8], style transfer [9], domain adaptation [10] and data augmentation [11].”
Class-Weighted Classification: Trade-offs and Robust Approaches (2020, methodology): “This has led to variants with different forms of data augmentation (Zhou & Liu, 2006; Mariani et al., 2018), but from a theoretical perspective, these methods remain poorly understood.”
Machine Learning for Synthetic Data Generation: a Review (2023, background): “The use of balanced synthetic datasets created by GANs to augment classification training has demonstrated the benefits for reducing disparate impact due to minoritized subgroup imbalance [196]–[198].”
LARGE: Latent-Based Regression through GAN Semantics (2021, background): “Others enriched a real dataset to improve performance in the low-data domain [4, 16, 59, 73] or on biased or unbalanced data [18,22,33,39,44].”
DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data (2021, methodology): “However, neither BAGAN nor GAMO focus on introducing artificial instances in a directed fashion to enhance class boundaries and improve the discrimination capabilities of a classifier trained on oversampled data.”
Artificial intelligence in the creative industries: a review (2020, methodology): “IBM Research introduced a Balancing GAN [305], where the model learns useful features from majority classes and uses these to generate images for minority classes that avoid features close to those of majority cases.”
Online continual learning with no task boundaries (2019, background): “However, this also adds the complexity that we need to train the GAN continually, which also depends on availability of task boundaries to keep balanced examples and avoid the mode collapse problem (Mariani et al., 2018).”