FTC Innovation Project · Prahova, Romania

Listening to the forest, so no species goes unheard.

Biovision is a network of three solar-powered smart nests that listen, watch and log — identifying bird species entirely on-device, then streaming every detection live over WiFi.

See how it works Component list
3nests across Prahova county
100%species ID on-device, not in the cloud
24/7continuous live monitoring
NEST-02 · Comarnic LIVE
Last detectionParus major
Confidence91.4%
Temperature18.6 °C
Humidity62 %RH
Battery / solar61% · charging
Storage used4.2 / 64 GB
ConnectionWiFi · stable
The problem

Biodiversity goes unmeasured wherever the road, the grid and the WiFi stop.

Most bird-monitoring setups assume a park ranger, a laptop and a signal. Real habitat rarely offers any of the three. Our brief was a distributed sensor network for Prahova county: three nests that watch nest conditions, identify visiting species entirely on-device, and stream every detection live over WiFi the moment it happens.

Streaming liveAll 3 nests broadcast detections over WiFi in real time
On-device AISpecies ID runs locally, not in the cloud
Field-readySolar-powered and weatherproofed to run unattended
How it works

One Raspberry Pi Zero 2 W per nest, listening year-round.

Each nest runs BirdNET-Pi, an open-source project from the Cornell Lab of Ornithology that recognizes bird species by song — entirely offline, on a single-board computer. Audio is more reliable and far cheaper to process than visual recognition from a low-cost camera trap, so it does the heavy lifting; motion, temperature and humidity sensing round out the picture. Every nest identifies species entirely on-device, then streams the result straight to the dashboard over WiFi.

Listen

An omnidirectional USB microphone feeds continuous audio into the nest's Raspberry Pi Zero 2 W.

Identify, on-device

BirdNET-Pi matches the recording against its species model locally — no connectivity needed, no audio leaves the nest.

Sense the nest

A DHT22 sensor logs temperature and humidity; an optional light sensor (LDR) correlates activity with the day/night cycle.

Trigger on motion

A PIR sensor wakes the Pi camera (or a cheaper standalone ESP32-CAM) only when something moves — saving both power and storage.

Log locally

Species, timestamp, temperature and humidity are written straight to the microSD card — the nest's memory until someone visits.

Stream, with local backup

Every detection streams to the Live Network below the moment it's logged. If WiFi drops, it's buffered on the microSD card and sent automatically once the connection returns.

Live network

Three nests, streaming in real time.

Where WiFi reaches the forest edge, Biovision skips the SD-card shuffle entirely. All three nests broadcast every detection the moment BirdNET-Pi logs it.

OpenStreetMap map of Prahova County showing the Biovision nest sites NEST-03 · Măneciu NEST-02 · Comarnic NEST-01 · Vălenii de Munte

Map data © OpenStreetMap contributors. Pin positions are approximate.

Live via WiFi Pulses on every new detection

Live feed — preview

LIVE

Sample stream. Real detections will appear here once the first live nest goes into the field.

NestSpeciesConfidenceTime
NEST-02 · ComarnicFringilla coelebs88%moments ago
NEST-01 · Vălenii de MunteTurdus merula94%1 min ago
NEST-03 · MăneciuErithacus rubecula90%3 min ago
NEST-02 · ComarnicPhylloscopus collybita82%6 min ago
NEST-01 · Vălenii de MunteDendrocopos major96%9 min ago
NEST-01 · Vălenii de MunteLIVE
Last detectionTurdus merula
Confidence94%
Battery / solar74% · charging
ConnectionWiFi · stable
NEST-02 · ComarnicLIVE
Last detectionParus major
Confidence91.4%
Battery / solar61% · charging
ConnectionWiFi · stable
NEST-03 · MăneciuLIVE
Last detectionErithacus rubecula
Confidence90%
Battery / solar68% · charging
ConnectionWiFi · stable
Identify a bird

Snap a bird? Find out what it is — and how rare.

Upload a photo you took yourself. Claude looks at it and reports the species, a rarity read for the region, and a short note — right here in your browser.

Checking availability…
Inside the nest

Component list.

ComponentNotes
Raspberry Pi Zero 2 WOfficial price $15 — check current price locally
microSD card, 32–64GB (A1/A2)A fast card is essential for BirdNET-Pi
Pi Camera (v2 or NoIR)NoIR useful for nighttime activity
USB microphoneA simple omnidirectional mic is enough
PIR sensorMotion trigger for the camera
DHT22 sensorTemperature + humidity
Battery (power bank / LiPo / 18650 + charger)Sized to the target autonomy
Small solar panel (5–10W) + charge controllerExtends field autonomy
Enclosure / nest body (wood, insulation, mounting)Waterproofing, ventilated sensor ports
Cables, connectors, misc. 
Power & field deployment

Built to run on sunlight, and to survive being rained on.

Validated on

Community testing confirms BirdNET-Pi runs acceptably on the Pi Zero 2 W under these conditions — see the GitHub discussion thread we used to validate the approach.

Roadmap

Next steps.

01

Single-nest prototype

Validate BirdNET-Pi, sensors and solar power on one nest before replicating across all 3.

02

Autonomy test

Run the prototype on battery/solar for a full week to measure real-world autonomy.

03

Collection protocol

Define how often nests are visited, by whom, and how memory cards get synced.

04

Live rollout

Bring the three WiFi-connected nests online and wire the dashboard to real telemetry.

05

Innovation Award narrative

Frame the real problem — biodiversity monitoring and bird conservation in Prahova — alongside the technical solution and its impact.

Team

Biovision is RO2D2's Innovation Project

FTC Team 17962 · Ploiești, Romania — representing "Mihai Viteazul" National College.

2018founded
15members
16volunteers