Enhanced Balancing GAN: Minority-class Image Generation

10/31/2020
by   Gaofeng Huang, et al.
6

Generative adversarial networks (GANs) are one of the most powerful generative models, but always require a large and balanced dataset to train. Traditional GANs are not applicable to generate minority-class images in a highly imbalanced dataset. Balancing GAN (BAGAN) is proposed to mitigate this problem, but it is unstable when images in different classes look similar, e.g. flowers and cells. In this work, we propose a supervised autoencoder with an intermediate embedding model to disperse the labeled latent vectors. With the improved autoencoder initialization, we also build an architecture of BAGAN with gradient penalty (BAGAN-GP). Our proposed model overcomes the unstable issue in original BAGAN and converges faster to high quality generations. Our model achieves high performance on the imbalanced scale-down version of MNIST Fashion, CIFAR-10, and one small-scale medical image dataset.

READ FULL TEXT

page 6

page 7

page 8

research
01/13/2022

Conditional Variational Autoencoder with Balanced Pre-training for Generative Adversarial Networks

Class imbalance occurs in many real-world applications, including image ...
research
03/26/2018

BAGAN: Data Augmentation with Balancing GAN

Image classification datasets are often imbalanced, characteristic that ...
research
06/17/2021

Class Balancing GAN with a Classifier in the Loop

Generative Adversarial Networks (GANs) have swiftly evolved to imitate i...
research
11/08/2017

Generalization of Deep Neural Networks for Chest Pathology Classification in X-Rays Using Generative Adversarial Networks

Medical datasets are often highly imbalanced, over representing common m...
research
05/14/2020

SAGE: Sequential Attribute Generator for Analyzing Glioblastomas using Limited Dataset

While deep learning approaches have shown remarkable performance in many...
research
03/22/2022

Dazzle: Using Optimized Generative Adversarial Networks to Address Security Data Class Imbalance Issue

Background: Machine learning techniques have been widely used and demons...
research
10/21/2020

One Model to Reconstruct Them All: A Novel Way to Use the Stochastic Noise in StyleGAN

Generative Adversarial Networks (GANs) have achieved state-of-the-art pe...

Please sign up or login with your details

Forgot password? Click here to reset