Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Computing | en_US |
| dc.contributor.advisor | Chung, Fu Lai Korris (COMP) | - |
| dc.creator | Lam, Raymond | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/10150 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | - |
| dc.rights | All rights reserved | en_US |
| dc.title | Study of the capability of generative adversarial network based models in image synthesis | en_US |
| dcterms.abstract | There has been having dramatically growth of research in Generative Adversarial Networks (GANs) since its introduction in 2014. GANs has been applied to various applications with impressive performance such as language processing and computer vision. Among those applications of GANs, image synthesis is perhaps the most popular and well-studied area including, but not limited to, image-to-image translation, image editing, image inpainting, and texture synthesis. GANs consist of two competitive deep neural networks called generator and discriminator. Due to the competitive training manner and the power of two deep neural networks, GANs are capable of producing reasonable and interesting images. It has already been demonstrated with great potential in image synthesis in different researches. GANs were only capable of generating black-and-white, blurry and small images in the past but now had been evolved to be capable of generating colorful, realistic and high-resolution images that are hardly be distinguished from the real one. In this thesis, we study and explore the image synthesis capability of Generative Adversarial Network (GAN) as motivated by its impressive but cherry pick performance. We proceed with experiments with taxonomy conditions of datasets provided, e.g. object colors, object sizes, and classes, so as to reveal how different datasets affect the results. The experiments include unpaired image-to-image translation model, Cycle-consistent Adversarial Networks (CycleGAN), and conditional image synthesis with Auxiliary Classifier GANs (ACGAN). We discuss the results and carry out the evaluations as well as point out the possible research directions of GAN in image synthesis. | en_US |
| dcterms.extent | ix, 64 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2019 | en_US |
| dcterms.educationalLevel | M.Sc. | en_US |
| dcterms.educationalLevel | All Master | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.LCSH | Machine learning | en_US |
| dcterms.LCSH | Neural networks (Computer science) | en_US |
| dcterms.LCSH | Image processing -- Digital techniques | en_US |
| dcterms.accessRights | restricted access | en_US |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 991022268444303411.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 9.05 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/10150