메뉴 건너뛰기
.. 내서재 .. 알림
소속 기관/학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색
질문

논문 기본 정보

자료유형
학술대회자료
저자정보
Jaewon Park (Korea Advanced Institute of Science and Technology) Changki Sung (Korea Advanced Institute of Science and Technology) Seunghee Lee (Korea Advanced Institute of Science and Technology) Dongwan Kang (Hanwha Aerospace) Hyun Myung (KI-Robotics, Korea Advanced Institute of Science and Technology)
저널정보
제어로봇시스템학회 제어로봇시스템학회 국제학술대회 논문집 ICCAS 2024
발행연도
2024.10
수록면
1,078 - 1,083 (6page)

이용수

표지
📌
연구주제
📖
연구배경
🔬
연구방법
🏆
연구결과
AI에게 요청하기
추천
검색
질문

초록· 키워드

오류제보하기
Cross-view geo-localization (CVGL) is the problem of determining the location of a ground-level camera with respect to an extensive database of geo-tagged aerial images. While GNSS can provide location data, using images alone for localization can be especially beneficial in scenarios where GNSS signals are obstructed or unreliable. Although CVGL is treated as a retrieval-based visual place recognition (VPR) task, it is more challenging due to the critical viewpoint domain gap. To reduce this gap, contrastive learning methods have demonstrated superior performance, especially by applying hard negative sampling. However, the issue with previous hard negative sampling is its high computational cost, making training time-consuming when performed at every epoch. In this study, we propose a more effective hard negative sampling method. We introduce sampling strategies that remove unnecessary similarity map calculations, effectively computing only for the queries. We utilize the proposed effective hard negative sampling method, conducting contrastive learning with the information noise-contrastive estimation (InfoNCE) loss function. Our work demonstrated more efficient training and superior performance compared with the existing approaches on CVUSA and VIGOR, common cross-view datasets.

목차

Abstract
1. INTRODUCTION
2. RELATED WORKS
3. METHODOLOGY
4. EXPERIMENTAL RESULT
5. CONCLUSION
REFERENCES

참고문헌 (0)

참고문헌 신청

함께 읽어보면 좋을 논문

논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!

최근 본 자료

전체보기

댓글(0)

0