Download the software
Everything we use to plan a survey, set up the aircraft, fly it in the simulator and look at the pictures afterwards. Our own code is free to use, change and share under the MIT licence; the parts made by others keep their licences and are credited below.
- a fixed-wing plane running ArduPlane (ArduPilot)
- a flight controller that can run Lua scripts (H7 or F7 boards such as the Matek H743)
- an airspeed sensor, GPS with compass and a battery monitor
- for recognition: a Raspberry Pi 4 and a network (RTSP) or USB camera; other cameras need a lens calibration
- a laptop with Windows, macOS or Linux
- quadcopters and other multirotors
- DJI, PX4 or iNav flight controllers
- small F4 boards that cannot run scripts
- planes without an airspeed sensor
Even on the right aircraft: answer every question about your plane in the Ground Station, fly the mission in its simulator first, check the control directions on the bench, and get your flight authorisation.
Plan, send, launch and look at the pictures, with a guided setup for a new aircraft. Install it like any program; nothing else needed.
Windows may say the publisher is unknown, because the app is not code-signed: choose More info, then Run anyway. On macOS, open it the first time with right-click, Open. Checksums: SHA256SUMS.txt on the release page. The source and the model on their own are attached to each release.
Version 2.3.3, built 2026-11-18 from commit 4c60b20ba6. Checksums of every file: SHA256SUMS.txt on the release page.
| You want to | Take |
|---|---|
| Plan and fly a survey, see the pictures | The app above. Its guided setup takes a new aircraft step by step to ready-to-fly; GROUND_STATION.md explains every tab. |
| Change the software, or run it from source | The source: on GitHub, or the software zip attached to each release. Start Start Ground Station.bat (Windows) or start_ground_station.sh; FLIGHT_READY.md describes the whole procedure. |
| Set up the Raspberry Pi 4 | Nothing to download: the app installs it over your Wi-Fi (setup step 7) and updates it later. |
| Only the flight controller settings | The app sends them itself. For Mission Planner, its Send tab saves the parameter file made for your aircraft (Full Parameter List, Load from file); the GPS-loss script is in the software zip (ardupilot/scripts). |
| Recognise tractors and livestock | Choose "Ready-made" on the Ground Station's Camera tab: the model is inside the app. On its own it is attached to each release. |
Read before flying: the software is tested in ArduPilot's simulator but not yet flown on a real aircraft. Fully automatic flight without a radio link needs the EASA "specific" category (an operational authorisation from Transpordiamet in Estonia), and taking pictures of people and property has its own rules. It comes as is, without any warranty.
The SHA-256 checksum of every file is in SHA256SUMS.txt on the release page. If the number you get is different, the file is damaged or not ours: do not use it, download it again.
| Computer | Command (in the folder with the file) |
|---|---|
| Windows | certutil -hashfile UAV-SILM-Ground-Station-Setup.exe SHA256 |
| macOS | shasum -a 256 UAV-SILM-Ground-Station-macos.zip |
| Linux | sha256sum -c SHA256SUMS.txt --ignore-missing |
Everything is built on GitHub from a published commit of our repository (the release page names it); the zips are built the same way every time, so anyone can rebuild them and get the same checksums: python3 onboard/tools/build_downloads.py --release --out DIR.
| Part | Made by | Licence |
|---|---|---|
| Our software, settings, script and documentation | UAV SILM team | MIT |
| Recognition model silm_oi | UAV SILM team, fine-tuned from Ultralytics YOLO11n (© Ultralytics) | AGPL-3.0 |
| Photos and boxes the model learnt from | Google Open Images; photos by their Flickr authors | CC BY 4.0 (boxes), CC BY 2.0 (photos) |
| Map in the Ground Station | Leaflet (Volodymyr Agafonkin and contributors) | BSD-2-Clause |
| Map data and tiles | © OpenStreetMap contributors | ODbL |
| Satellite imagery | © Esri and its data providers | Esri terms of use |
| Fonts (on this site and in the Ground Station) | Bricolage Grotesque Project Authors; IBM Corp. (IBM Plex Mono) | SIL Open Font License 1.1 (text, text) |
| Flight controller firmware and simulator (not included) | ArduPilot project | GPL-3.0 |
| Standard recognition model (downloaded on the Pi) | TensorFlow authors, COCO SSD MobileNet v1 | Apache-2.0 |
| Python libraries (installed separately) | pymavlink, OpenCV, NumPy, LiteRT, PyYAML, pySerial and others | LGPL-3.0, Apache-2.0, BSD, MIT |
The full list, with every author and licence text, is in THIRD_PARTY_NOTICES.md, in the app's folder and in the source. The four test photos in it are credited by name in tests/data/oi_samples/samples.json.
SILM Buteo is in development and has not flown yet. Figures are design estimates from CFD and calculation, not flight-test data. About the student project