核心概念
Proposing shifting-dense partition learning (SDPL) for accurate cross-view geo-localization against position deviations and scale changes.
摘要
The content introduces the shifting-dense partition learning (SDPL) framework for UAV-view geo-localization. It addresses challenges in matching images from different platforms by introducing dense partition and shifting-fusion strategies. Extensive experiments show competitive performance on University-1652 and SUES-200 datasets. The article includes an introduction, related works, detailed methods, experimental results, and conclusions.
- Introduction to UAVs and geo-localization.
- Challenges in cross-view geo-localization.
- Introduction of shifting-dense partition learning (SDPL).
- Detailed explanation of dense partition and shifting-fusion strategies.
- Experimental results on University-1652 and SUES-200 datasets.
- Conclusion and future directions.
統計資料
"Extensive experiments show that our SDPL is robust to position shifting and scale variations."
"Our SDPL achieves state-of-the-art performance on University-1652 and SUES-200."
引述
"We propose a shifting-dense partition learning (SDPL) to achieve accurate cross-view geo-localization against position deviations and scale changes."
"Extensive experiments show that our SDPL achieves state-of-the-art performance on two public datasets, i.e., University-1652 and SUES-200."