YILDIZ-VPR: A New Dataset for Visual Place Recognition with Dense, Diverse Pedestrian-Level Imagery
At Yildiz Technical University, researchers have developed a novel dataset named YILDIZ-VPR to advance Visual Place Recognition (VPR). This dataset was compiled by conducting multiple walking sessions across the Davutpasa campus, capturing a variety of images under different lighting, seasonal, and weather conditions, showcasing both historical structures and green spaces. Each video was filmed using a GoPro 9 camera, with GPS data synchronized and supplemented by additional sensor metrics such as gyroscope and speed information. The primary goal of this dataset is to facilitate geo-localization research, as detailed in a paper available on arXiv (identifier 2608.17033). YILDIZ-VPR emphasizes pedestrian perspectives and extensive coverage, potentially enhancing place recognition for navigation in real-world scenarios.
Key facts
- The dataset is named YILDIZ-VPR.
- It is designed for Visual Place Recognition (VPR).
- Data was collected on the Davutpasa campus of Yildiz Technical University.
- Collection involved repeated walking traversals.
- Images capture outdoor scenes at different times of day, seasons, and weather conditions.
- Content includes historical buildings, modern structures, roads, green areas, and wooded regions.
- Videos were recorded with a GoPro 9 camera.
- Frames are synchronized with GPS and auxiliary sensor data including gyroscope and speed.
Entities
Institutions
- Yildiz Technical University
Locations
- Davutpasa campus
- Yildiz Technical University