Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network
The processing of facial images is an important task, because it is required for a large number of real-world applications. As deep-learning models evolve, they require a huge number of images for training. In reality, however, the number of images available is limited. Generative adversarial networ...
Main Authors: | , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-03-01
|
Series: | Applied Sciences |
Subjects: | |
Online Access: | https://www.mdpi.com/2076-3417/10/6/1995 |
id |
doaj-c19457024a28497ca28589ebd904297f |
---|---|
record_format |
Article |
spelling |
doaj-c19457024a28497ca28589ebd904297f2020-11-25T01:30:04ZengMDPI AGApplied Sciences2076-34172020-03-01106199510.3390/app10061995app10061995Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial NetworkJeong gi Kwak0Hanseok Ko1School of Electrical Engineering, Korea University, Seoul 136-701, KoreaSchool of Electrical Engineering, Korea University, Seoul 136-701, KoreaThe processing of facial images is an important task, because it is required for a large number of real-world applications. As deep-learning models evolve, they require a huge number of images for training. In reality, however, the number of images available is limited. Generative adversarial networks (GANs) have thus been utilized for database augmentation, but they suffer from unstable training, low visual quality, and a lack of diversity. In this paper, we propose an auto-encoder-based GAN with an enhanced network structure and training scheme for Database (DB) augmentation and image synthesis. Our generator and decoder are divided into two separate modules that each take input vectors for low-level and high-level features; these input vectors affect all layers within the generator and decoder. The effectiveness of the proposed method is demonstrated by comparing it with baseline methods. In addition, we introduce a new scheme that can combine two existing images without the need for extra networks based on the auto-encoder structure of the discriminator in our model. We add a novel double-constraint loss to make the encoded latent vectors equal to the input vectors.https://www.mdpi.com/2076-3417/10/6/1995generative modelsgan (generative adversarial networks)facial imagegenerationdatabase augmentationsynthesis |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Jeong gi Kwak Hanseok Ko |
spellingShingle |
Jeong gi Kwak Hanseok Ko Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network Applied Sciences generative models gan (generative adversarial networks) facial image generation database augmentation synthesis |
author_facet |
Jeong gi Kwak Hanseok Ko |
author_sort |
Jeong gi Kwak |
title |
Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network |
title_short |
Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network |
title_full |
Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network |
title_fullStr |
Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network |
title_full_unstemmed |
Unsupervised Generation and Synthesis of Facial Images via an Auto-Encoder-Based Deep Generative Adversarial Network |
title_sort |
unsupervised generation and synthesis of facial images via an auto-encoder-based deep generative adversarial network |
publisher |
MDPI AG |
series |
Applied Sciences |
issn |
2076-3417 |
publishDate |
2020-03-01 |
description |
The processing of facial images is an important task, because it is required for a large number of real-world applications. As deep-learning models evolve, they require a huge number of images for training. In reality, however, the number of images available is limited. Generative adversarial networks (GANs) have thus been utilized for database augmentation, but they suffer from unstable training, low visual quality, and a lack of diversity. In this paper, we propose an auto-encoder-based GAN with an enhanced network structure and training scheme for Database (DB) augmentation and image synthesis. Our generator and decoder are divided into two separate modules that each take input vectors for low-level and high-level features; these input vectors affect all layers within the generator and decoder. The effectiveness of the proposed method is demonstrated by comparing it with baseline methods. In addition, we introduce a new scheme that can combine two existing images without the need for extra networks based on the auto-encoder structure of the discriminator in our model. We add a novel double-constraint loss to make the encoded latent vectors equal to the input vectors. |
topic |
generative models gan (generative adversarial networks) facial image generation database augmentation synthesis |
url |
https://www.mdpi.com/2076-3417/10/6/1995 |
work_keys_str_mv |
AT jeonggikwak unsupervisedgenerationandsynthesisoffacialimagesviaanautoencoderbaseddeepgenerativeadversarialnetwork AT hanseokko unsupervisedgenerationandsynthesisoffacialimagesviaanautoencoderbaseddeepgenerativeadversarialnetwork |
_version_ |
1725093879575740416 |