DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Choi, Jaesik | - |
dc.contributor.advisor | 최재식 | - |
dc.contributor.author | Choi, Hwanil | - |
dc.date.accessioned | 2023-06-22T19:31:28Z | - |
dc.date.available | 2023-06-22T19:31:28Z | - |
dc.date.issued | 2022 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1008218&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/308228 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2022.8,[iv, 24 p. :] | - |
dc.description.abstract | Even though Generative Adversarial Networks (GANs) have shown a remarkable ability to generate high- quality images, GANs do not always guarantee the generation of photorealistic images. Occasionally, they generate images that have defective or unnatural objects, which are referred to as ‘artifacts’. Research to investigate why these artifacts emerge and how they can be detected and removed has yet to be sufficiently carried out. To analyze this, we first hypothesize that rarely activated neurons and frequently activated neurons have different purposes and responsibilities for the progress of generating images. In this study, by analyzing the statistics and the roles for those neurons, we empirically show that rarely activated neurons are related to the failure results of making diverse objects and inducing artifacts. In addition, we suggest a correction method, called ‘Sequential Ablation’, to repair the defective part of the generated images without high computational cost and manual efforts. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | generative adversarial networks▼aexplainable ai▼aimage correction | - |
dc.subject | 적대적 생성 네트워크▼a설명가능 인공지능▼a이미지 수리 | - |
dc.title | Can we find neurons that cause unrealistic images in deep generative networks? | - |
dc.title.alternative | 생성모델의 비정상적 이미지 생성에 대한 원인 분석과 탐지 및 수리 기법 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :김재철AI대학원, | - |
dc.contributor.alternativeauthor | 최환일 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.