README.md

---
license: mit
task_categories:
- depth-estimation
- keypoint-detection
pretty_name: AerialExtreMatch Localization
viewer: false
tags:
- image
- geospatial
---
# AerialExtreMatch — Localization Dataset

[Code](https://github.com/Xecades/AerialExtreMatch) | [Project Page](https://xecades.github.io/AerialExtreMatch/) | Paper (WIP)

This repo contains the **localization** set for our paper *AerialExtreMatch: A Benchmark for Extreme-View Image Matching and Localization*. Two different quality map and 264 query images are included. We also provide [**benchmark**](https://huggingface.co/datasets/Xecades/AerialExtreMatch-Benchmark) and [**train**](https://huggingface.co/datasets/Xecades/AerialExtreMatch-Train) datasets.

## Usage

Simply clone this repository.

```bash
git clone [email protected]:datasets/Xecades/AerialExtreMatch-Train
```

## Dataset Structure

<pre>
.
├── HQref: *.tif
├── LQref: *.tif,*.prj,*.tfw
├── rgb: *.JPG
├── gt_pose.txt
├── intrinsic.txt <i>(see below)</i>
└── pose.txt <i>(see below)</i>
</pre>

 - `pose.txt` provides the noisy prior UAV pose in the format: `name.JPG wq wx wy wz x y z`.
 - `intrinsic.txt` provides the intrinsic parameters of the UAV camera. Only undistorted images and parameters are provided. Format: `name.JPG PINHOLE width height fx fy cx cy`.