Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks

碩士 === 國立臺灣科技大學 === 電機工程系 === 106 === We present a conditional high resolution image generating system that can synthesize face images based on input of a wide range of facial attributes. First we focus on the training and optimization of an image generation model that synthe sizes high resolution i...

Full description

Bibliographic Details
Main Authors: Ting-Jia Yang, 楊庭嘉
Other Authors: Nai-Jian Wang
Format: Others
Language:en_US
Published: 2018
Online Access:http://ndltd.ncl.edu.tw/handle/my7yyw
id ndltd-TW-106NTUS5442086
record_format oai_dc
spelling ndltd-TW-106NTUS54420862019-05-16T00:59:40Z http://ndltd.ncl.edu.tw/handle/my7yyw Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks 基於生成對抗網路之條件式高解析度圖像生成 Ting-Jia Yang 楊庭嘉 碩士 國立臺灣科技大學 電機工程系 106 We present a conditional high resolution image generating system that can synthesize face images based on input of a wide range of facial attributes. First we focus on the training and optimization of an image generation model that synthe sizes high resolution images based on the Wasserstein distance and progressive growing method. Then to exert control over the image generation model, we deviate from the current trend and instead explore the viability of augmenting the image generation model by introducing a second generative adversarial network that learns its conditional latent space distribution in hope to manipulate the attributes of its output images. Experiments were conducted on CelebA and MNIST dataset. Results show that our method can convert an existing generative model to a conditional generative model that synthesize images based on input class or attributes while retaining reasonable diversity. Furthermore, this method allows one to incorporate labels from multiple datasets without retraining the generative model from scratch. Nai-Jian Wang 王乃堅 2018 學位論文 ; thesis 49 en_US
collection NDLTD
language en_US
format Others
sources NDLTD
description 碩士 === 國立臺灣科技大學 === 電機工程系 === 106 === We present a conditional high resolution image generating system that can synthesize face images based on input of a wide range of facial attributes. First we focus on the training and optimization of an image generation model that synthe sizes high resolution images based on the Wasserstein distance and progressive growing method. Then to exert control over the image generation model, we deviate from the current trend and instead explore the viability of augmenting the image generation model by introducing a second generative adversarial network that learns its conditional latent space distribution in hope to manipulate the attributes of its output images. Experiments were conducted on CelebA and MNIST dataset. Results show that our method can convert an existing generative model to a conditional generative model that synthesize images based on input class or attributes while retaining reasonable diversity. Furthermore, this method allows one to incorporate labels from multiple datasets without retraining the generative model from scratch.
author2 Nai-Jian Wang
author_facet Nai-Jian Wang
Ting-Jia Yang
楊庭嘉
author Ting-Jia Yang
楊庭嘉
spellingShingle Ting-Jia Yang
楊庭嘉
Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks
author_sort Ting-Jia Yang
title Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks
title_short Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks
title_full Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks
title_fullStr Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks
title_full_unstemmed Conditional High-Resolution Image Synthesizing Based on Generative Adversarial Networks
title_sort conditional high-resolution image synthesizing based on generative adversarial networks
publishDate 2018
url http://ndltd.ncl.edu.tw/handle/my7yyw
work_keys_str_mv AT tingjiayang conditionalhighresolutionimagesynthesizingbasedongenerativeadversarialnetworks
AT yángtíngjiā conditionalhighresolutionimagesynthesizingbasedongenerativeadversarialnetworks
AT tingjiayang jīyúshēngchéngduìkàngwǎnglùzhītiáojiànshìgāojiěxīdùtúxiàngshēngchéng
AT yángtíngjiā jīyúshēngchéngduìkàngwǎnglùzhītiáojiànshìgāojiěxīdùtúxiàngshēngchéng
_version_ 1719172422583713792