Repository at Hanyang UniversityGRADUATE SCHOOL[S](대학원)ARTIFICIAL INTELLIGENCE(인공지능학과)Theses(Master)
100
0
Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | 김태현 | - |
| dc.contributor.author | 임은우 | - |
| dc.date.accessioned | 2024-03-01T07:32:22Z | - |
| dc.date.available | 2024-03-01T07:32:22Z | - |
| dc.date.issued | 2023.8 | - |
| dc.identifier.uri | http://hanyang.dcollection.net/common/orgView/200000684623 | en_US |
| dc.identifier.uri | https://repository.hanyang.ac.kr/handle/20.500.11754/188217 | - |
| dc.description.abstract | Recentresearchhasoftenreliedontherepresentationpowerofneuralnetworksandoverlookedseveralfactorsinvolvedinhazedegradation,includingtransmissionandatmosphericlight.Ingeneral,thesefactorsareunknown,resultingininherentuncertainties.Inthisstudy,weintroduceavariationalBayesianframeworkforsingleimagedehazingtoaddresstheseuncertaintiesandaccountforthefactorsinvolvedinhazedegradation.Consideringkeycomponentsofhazedegradationaslatentvariables,theposteriordistributionsareparameterizedbycorrespondingtwobranchesofneuralnetworks,respectively.Basedontheatmosphericscatteringmodel,theproposedframeworkresultsinanewobjectivefunctionthatenablescooperationbetweenthesebranchesbyjointoptimizationandleadstoanamplificationoftheperformanceofeachother.Furthermore,themodel-agnosticframeworkfacilitatesnotonlyeasyadaptationofanyexistingdehazingnetworkswithoutmodificationofarchitecturebutalsonoextraoverheadintheinferencephase.Extensiveexperimentshavedemonstratedconsistentimprovementsintheperformanceofbaselinemethodsacrossdifferentdatasetsandmodels. | - |
| dc.publisher | 한양대학교 | - |
| dc.title | DeepVariationalFrameworkforSingleImageDehazing | - |
| dc.type | Theses | - |
| dc.contributor.googleauthor | 임은우 | - |
| dc.sector.campus | S | - |
| dc.sector.daehak | 대학원 | - |
| dc.sector.department | 인공지능학과 | - |
| dc.description.degree | Master | - |
- Appears in Collections:
- GRADUATE SCHOOL[S](대학원) > ARTIFICIAL INTELLIGENCE(인공지능학과) > Theses(Master)
- Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
