fugleramme

E-ink bird frame for Raspberry Pi - real-time bird detection by audio, fully local AI, rendered as real, hand-cut 1800s bird illustrations.

Sorry about the dirty window - squirrels have been stealing the bird food.

[!NOTE] Still in early development: expect the odd bug and a few unpolished edges, with plenty more features to come.

Live on fugleramme.arnegiacomo.dev running from my kitchen window and displaying the actual birds currently heard in my garden (Bergen, Norway).

Hardware, install and operations docs: arnegiacomo.dev/fugleramme

How it works

BirdNET-Go listens on a mic and handles the classifier. Fugleramme polls its api, matches each species to an illustration, then packs them onto a page, and redraws only when the birds change - on an Inky Impression e-ink panel, and as a web kiosk serving the same view. There's an admin page that lets you configure what to show, and automatic updates and such.

If you already run BirdNET-Go, point the frame at it instead - on the same machine or anywhere else reachable from your network.

[!TIP] The e-ink panel is not required, although it's recommended for the intended experience. Without one, Fugleramme runs web-only - show the kiosk on a display over HDMI, or open it from any device on the network.

Hardware

A Raspberry Pi 5, an Inky Impression 13.3" (Spectra 6), a mic and an A4 frame. Full parts list, recommendations and alternatives: Hardware.

Art

Half the point of this project is showing off some amazing public-domain natural-history illustrations. Over 800 cut-outs covering more than 400 species, every one taken from a real plate and hand-curated for this project (no art is AI-generated, though some has been retouched with AI).

Each detected species is matched to its illustration, background-removed, and packed onto a textured paper page with the larger birds toward the centre, sized by body mass. An empty window shows a bare perch.

The plates are Scandinavian, British and central European, so the Nordics, the British Isles and Germany are best covered. Elsewhere not so much (yet). Broader European and North American coverage is in the works!

See Adding artwork for manual cutout steps.

No detections	A few visitors	A full garden

Run locally (for development)

uv sync                                       # set up venv
uv run fugleramme-fake-detector               # stand-in BirdNET-Go on :8090
uv run fugleramme-dev                         # start service on :8080 with hot-reload

The fake detector's flags, and working against a real station instead: Running it without a Pi.

Install on a Raspberry Pi

From the pi (assuming you have the hardware up and running):

curl -fsSL https://raw.githubusercontent.com/arnegiacomo/fugleramme/main/install.sh | bash

Asks where BirdNET-Go should live and which ports to use, clones the repo, installs the required deps, and starts the frame as a systemd service. NB! Will probably require a reboot on a fresh system.

From a blank SD card, see the full install guide.

Run in a container

docker run -d -p 8080:8080 -v fugleramme:/data \
  -e FUGLERAMME_DETECTOR_URL=http://birdnet.local:8080 \
  ghcr.io/arnegiacomo/fugleramme

Or build the image from a checkout:

docker build -t fugleramme .
docker run --rm -p 8080:8080 -v fugleramme:/data \
  -e FUGLERAMME_DETECTOR_URL=http://birdnet.local:8080 fugleramme

Kiosk on :8080, admin on :8080/admin, everything it persists in /data.

On a Linux box with a USB mic, this brings up BirdNET-Go alongside it:

curl -fsSL https://raw.githubusercontent.com/arnegiacomo/fugleramme/main/examples/docker-compose.yml -o docker-compose.yml
docker compose up -d

See Container for more info.

Contributing

Contributions are very welcome and encouraged - fixes, docs and artwork most of all. Thanks to everyone who has contributed so far ❤️

• Something is broken - a bug report
• A question, an idea, or a frame you have built - the FAQ first, then Discussions
• A fix, a doc change, or a bird you have cut - open a PR, no issue needed

See Contributing for more info.

License

• Code: MIT - see LICENSE.
• Detection (BirdNET-Go, installed separately as a container): CC BY-NC-SA 4.0, non-commercial only. BirdNET model by the Cornell Lab of Ornithology and Chemnitz University of Technology, taxonomy data powered by eBird.org.
• Bird images: each style folder carries its own terms and sources, and its manifest links the plate every file was cut from. classic is CC BY-SA 4.0 - see assets/artwork/classic/ATTRIBUTION.md.
• Label fonts (assets/fonts/): SIL OFL 1.1 - see assets/fonts/ATTRIBUTION.md.
• Bird sizes (assets/bird_sizes.csv): body mass from AVONET (Tobias et al. 2022, Ecology Letters, doi:10.1111/ele.13898), CC BY 4.0.
• BirdNET scientific-name aliases (assets/birdnet_aliases.json): OpenFauna's compiled taxonomic alias map, CC BY-SA 4.0 - see assets/ATTRIBUTION.md.

Prebuilt frames

I've built a few of these. If you'd like one rather than building it yourself, please get in touch.