# Bad Wurzach 2023 – drone dataset (abandoned farmstead / fire ruin)

Open photogrammetry dataset published by **Bitblade** (https://bitblade.io) under **CC BY 4.0**.

| | |
|---|---|
| Images | 1 471 JPEG (single camera) |
| Total size | ~17 GB |
| Camera | DJI Mavic 2 Pro – Hasselblad L1D-20c, 5472 × 3648 px |
| Date | 18 May 2023 |
| Location | Bad Wurzach area, Baden-Württemberg, Germany |
| Georeference | GPS + DJI XMP (gimbal angles, relative altitude) in every image; no GCPs, no RTK |
| Capture | Mixed nadir and oblique flights (~21 m and ~87 m above take-off) plus low facade shots of a residential house, a fire ruin and sheds; 3 images (DJI100_0392–0394) look at the horizon |

## Privacy / anonymisation

Before publication all images were anonymised; please keep it that way in derived works:

- licence plates (detector + manual review),
- faces and the head region of every detected person,
- **every vehicle larger than ~120 px is pixelated completely** (cars, van, trailers, tractor).

Changed images were re-encoded as JPEG (quality 95); all EXIF/XMP metadata (GPS, DJI gimbal tags) were copied from the originals. Images without findings are byte-identical to the originals.

## Processing notes (ODM 3.6)

The dataset is a good test case for mixed nadir/oblique/facade captures:

- Default settings produced a model with flipped Z axis ("Negative GSD") and DensifyPointCloud failures.
- What worked for us (1 453 / 1 468 images in one reconstruction, reprojection error 0.83 px, GSD 0.60 cm, mean GPS deviation 1.69 m): keep the facade/oblique images (they stabilise the model), exclude only the 3 horizon images, OpenSfM `align_method: orientation_prior`, incremental SfM, no auto-boundary.

## Files

- `images/` – the 1 471 images (download individually, list in `images.txt`)
- `bad-wurzach-2023-images.zip` – all images in one archive (uncompressed zip)
- `SHA256SUMS` – checksums of all images
- `LICENSE.txt` – CC BY 4.0

## Licence and attribution

Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/

Please credit: **"Bad Wurzach 2023 drone dataset © Bitblade (bitblade.io), CC BY 4.0"**.