- About 70km flight data with various terrian, multi-height and illumination changes.
- 11 flight sequences with different routes ranging from the shortest one of 3.7km to the longest up to 11km.
- 18361 separate aerial-based images with 14096 cropped corresponding map patches.
- Adequate evaluations on VPR, VAL and VO to serve as the baselines for aerial vehicles.
- All samples include high-res images, IMU, and GNSS tag.
| Sensor | Details |
|---|---|
| Camera | FLIR BFS-U3-31S4C-C; RGB channels, 2048×1536 resolution, 55Hz max frame rate (20Hz in experiment), global shutter |
| Lens | Chiopt FA0401C; 82.9 horizontal FOV, 66.5 vertical FOV, 4∼75mm focal length |
| IMU | Xsens MTi-30-2A8G4; 200Hz update with acceleration and angular velocity |
| GNSS | NovAtel OEM718D; dual-antenna, 5Hz update, 1.5m (RMS) with single point, 1cm + 1ppm (RMS) with RTK |
| Altimeter | Benewake TF350; 10Hz update, 350m maximum detection range with 0.1m accuracy |
Code for VPR is mainly inspired by the evaulations given in DVGL Benchmark. And we add some re-rank methods and re-implement it for the AerialVL dataset. The full code is coming soon.
The paper as well as the full dataset will be released after the review process completing.
GPLv3 license.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent