DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kim, Junmo | - |
dc.contributor.advisor | 김준모 | - |
dc.contributor.author | Kim, Doyeon | - |
dc.date.accessioned | 2023-06-23T19:33:40Z | - |
dc.date.available | 2023-06-23T19:33:40Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1030547&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/309093 | - |
dc.description | 학위논문(박사) - 한국과학기술원 : 전기및전자공학부, 2023.2,[v, 49 p. :] | - |
dc.description.abstract | This paper aims to bridge the gap between deep learning technology and visual content creation by proposing several methods for generating images and videos based on various types of user input through generative adversarial networks. Recently, deep neural networks have shown improved performance and rapid development in many fields. Specifically, generative adversarial networks that utilize the adversarial learning of two neural networks, a generator and a discriminator, produce very diverse and realistic result images. This work mainly consists of two major generative tasks. First, we propose a conditional video learning methodology based on adversarial neural networks. Second, we suggest new approaches to transferring real-world images and videos to the artistic domain. We show the effectiveness of the proposed methods by conducting extensive qualitative and quantitative experiments on diverse datasets. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Deep learning▼aMachine learning▼aComputer vision▼aGenerative model▼aGenerative adversarial networks | - |
dc.subject | 딥러닝▼a머신 러닝▼a컴퓨터 비전▼a생성 모델▼aGAN | - |
dc.title | (A) study on the creation of various images and videos based on user input using deep learning networks | - |
dc.title.alternative | 딥러닝 모델을 이용한 유저 입력 기반의 다양한 이미지 및 비디오 생성에 대한 연구 | - |
dc.type | Thesis(Ph.D) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | 김도연 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.