DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Bae, Hyeon-Min | - |
dc.contributor.advisor | 배현민 | - |
dc.contributor.author | Lee, Seunghyun | - |
dc.date.accessioned | 2022-04-27T19:31:19Z | - |
dc.date.available | 2022-04-27T19:31:19Z | - |
dc.date.issued | 2021 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=963432&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/296000 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전기및전자공학부, 2021.8,[iv, 28 p. :] | - |
dc.description.abstract | This paper deals with a neural network that mimics biological structure of bat so that 3-dimensional (3D) environments can be perceived through fusion of auditory and visual information, named as Bat-FNet. The autonomous vehicles typically use visual sensors such as RADAR, LIDAR, and RGB cameras, and sound sensor like ultrasonic. Visual sensors are vulnerable to adverse weather, where sight is not secured. Ultrasonic sensors are used only for measuring distance even though they are robust [54]. The Bat-FNet, inspired by bats that use eyes and ears harmoniously to survive in complex environments, recognizes location and size of the target object. We prove the superiority of fusion network via mean square error (MSE) and intersection over union (IoU) scores. We demonstrate robustness against image distortion by complementing each other between ultrasound and camera sensors. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Ultrasound▼abat▼aautonomous vehicle▼aneural network▼acamera | - |
dc.subject | 초음파▼a박쥐▼a자율주행▼a뉴럴 네트워크▼a카메라 | - |
dc.title | Bat-FNet : bat-inspired audiovisual fusion neural network for 3d information reconstruction | - |
dc.title.alternative | 3차원 정보 추론을 위한 박쥐 모방 시청각 융합 신경망 개발 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | 이승현 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.