DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 노용만 | - |
dc.contributor.author | Han, Jongwook | - |
dc.contributor.author | 한종욱 | - |
dc.date.accessioned | 2024-07-25T19:31:14Z | - |
dc.date.available | 2024-07-25T19:31:14Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1045903&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/320673 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전기및전자공학부, 2023.8,[iii, 20 p. :] | - |
dc.description.abstract | Deep learning-based pedestrian detectors are now being used in various applications, including surveillance cameras and autonomous vehicles. However, the lack of generalizability of pedestrian detectors remains a problem. Recently, it has been shown that utilizing the knowledge of large-scale models on pedestrian detection can improve the generalizability of pedestrian detectors. However, the current method uses only a single pedestrian dataset to extract pedestrian knowledge from a large-scale model. In this paper, we propose a data curation method to gather clean and diverse pedestrian instances from multiple pedestrian datasets. To filter noisy pedestrian instances, we propose CLIP-based Pedestrian Filtering Module (CPFM). CPFM utilizes the image-text-aligned property of CLIP model to filter noisy pedestrian instances. Through extensive experiments on various pedestrian datasets, we show the effectiveness and the generalizability of our proposed method. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | 보행자 검출▼a대규모 모델▼a지식 추출 | - |
dc.subject | Pedestrian detection▼aLarge-scale model▼aKnowledge extraction | - |
dc.title | Extracting pedestrian knowledge from curated multi-dataset with large-scale model | - |
dc.title.alternative | 데이터 큐레이션과 대규모 모델을 활용한 다중 데이터셋으로부터의 보행자 지식 추출 방법 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | Ro, Yong Man | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.