DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 양은호 | - |
dc.contributor.author | Shin, Jonghoon | - |
dc.contributor.author | 신종훈 | - |
dc.date.accessioned | 2024-07-25T19:30:46Z | - |
dc.date.available | 2024-07-25T19:30:46Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1045726&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/320538 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2023.8,[iv, 22 p. :] | - |
dc.description.abstract | Motion Transfer Model animates an image by capturing motion from a driving video. In most motion transfer models, unsupervised learning is commonly employed, which neglects the consideration of segment information in the images and results in motion estimation without segment awareness. This negatively impacts each module of the motion transfer model and ultimately affects the quality of the generated videos. In this paper, we analyze the impact on each motion transfer module when segment information is not taken into account and propose \emph{Segment-Aware Notion Transfer Model} that utilizes segmentation maps for motion transfer. We validate our approach on benchmark dataset and observe improvements in the image quality of the genereated video, both on quantitatively and qualitatively. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | 움직임 전이▼a영상 생성▼a사진 분활 정보 | - |
dc.subject | Motion transfer▼aVideo generation▼aSegmentation map | - |
dc.title | Segment-aware motion model for image animation | - |
dc.title.alternative | 사진 분활 정보를 활용한 움직임 전이 신경망 모델 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :김재철AI대학원, | - |
dc.contributor.alternativeauthor | Yang, Eunho | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.