EstUAV

Tooling for processing DJI Mavic 3T drone flight data into datasets suitable for visual geolocation pipelines such as WILDNAV.

Raw datasets are published on Hugging Face: https://huggingface.co/datasets/Gekd/EstUAV.

Dataset Notes

The Hugging Face dataset ships two subsets. Use the matching --timezone value when processing each:

Hugging Face subset	Required --timezone
snow/	2
no_snow/	3

Dependencies

System libraries (install via your package manager or the upstream guides):

• libcurl
• OpenCV
• PROJ

Vendored dependencies (included under extern/): CLI11 and nlohmann/json.

Data Layout

Place input data under data/ at the repository root:

1. Copy the contents of the Mavic 3T's /DCIM directory into data/.
2. Add the flight .csv log file(s) to data/.

DJI log files are stored on the remote controller and are encrypted on firmware v13 and above. Extracting .csv files from the CLI requires a DJI API key; otherwise an online log viewer can be used.

Processed results are written to output/.

Build

mkdir build && cd build
cmake ..
make

Usage

./estuav --timezone <offset> [flags]

Flag	Description
-z, --timezone N	Timezone offset (required)
-w, --wildnav	Produce WILDNAV-formatted output
-r, --rgb	Include RGB images in output
-t, --thermal	Include thermal images in output

Example

./estuav -w -z 3 -r -t