DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kweon, In So | - |
dc.contributor.advisor | 권인소 | - |
dc.contributor.author | Jung, Yunjae | - |
dc.date.accessioned | 2021-05-11T19:31:43Z | - |
dc.date.available | 2021-05-11T19:31:43Z | - |
dc.date.issued | 2019 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=875223&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/282950 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 미래자동차학제전공, 2019.8,[v, 33 p. :] | - |
dc.description.abstract | In this thesis, we address the problem of unsupervised video summarization that automatically extracts key-shots from an input video. Specifically, we tackle two critical issues based on our empirical observations: (i) Ineffective feature learning due to flat distributions of output importance scores for each frame, and (ii) training difficulty when dealing with long-length video inputs. To alleviate the first problem, we propose a simple yet effective regularization loss term called variance loss. The proposed variance loss allows a network to predict output scores for each frame with high discrepancy which enables effective feature learning and significantly improves model performance. For the second problem, we design a novel two-stream network named Chunk and Stride Network (CSNet) that utilizes local (chunk) and global (stride) temporal view on the video features. Our CSNet gives better summarization results for long-length videos compared to the existing methods. In addition, we introduce an attention mechanism to handle the dynamic information in videos. This attention uses differences between adjacent frames in feature space. We demonstrate the effectiveness of the proposed methods by conducting extensive ablation studies and show that our final model achieves new state-of-the-art results on two benchmark datasets. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Unsupervised learning▼adeep learning▼avideo summarization▼acomputer vision▼afeature learning | - |
dc.subject | 비교사 학습▼a딥러닝▼a비디오 요약▼a컴퓨터 비전▼a특징 학습 | - |
dc.title | Discriminative feature learning for unsupervised video summarization | - |
dc.title.alternative | 비교사 학습을 이용한 비디오 요약을 위한 구별적 특징 학습 방법 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :미래자동차학제전공, | - |
dc.contributor.alternativeauthor | 정윤재 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.