1.6 million street-level images organized into sequences across 30 cities on 6 continents. The dataset includes metadata such as GPS coordinates, compass headings, and timestamps to support visual place recognition and localization research.
Use Cases
- Train visual place recognition models using GPS coordinates and image sequences to identify geographic locations
- Benchmark image retrieval algorithms against seasonal and lighting variations using the timestamp and city metadata
- Develop sequence-based localization systems that utilize the temporal ordering of frames
Strengths
- 1.6 million images spanning 30 cities across 6 continents
- Includes GPS coordinates, compass angles, and capture timestamps for every image
- Organized into sequences to enable temporal and motion-based localization analysis