The Songs of the Finnish Forest
Real-time bird song identification from a remote cottage in Finland
Today at the Cottage
9 Sept 2026 • 16:00 local time
Current conditions
14°C
few clouds
Late summer — Activity is high, but song is low. Birds are busy eating as much as possible to fuel up for the long flight south.
Who's Singing Right Now?
Latest bird detections, captured by our AI-powered audio monitoring system
Years of Birdsong
Patterns and insights from continuous monitoring across all seasons
Seasonal Migration
Total daily detections over the year showing spring arrival, peak breeding season, and autumn departure
Bird activity peaks dramatically in April–May as migratory species arrive for breeding season, then declines in August as they depart
Species Spotlight
Different migration patterns: year-round residents vs. seasonal visitors
Some species like the Common Raven stay year-round, while others like Common Chiffchaff visit only during summer months
Daily Rhythm
24-hour activity pattern showing the dawn chorus and evening peak
Birds sing most actively in the hour after sunrise (dawn chorus), with a smaller peak in the evening
Most Detected Species
Top 10 bird species by total number of detections
How It Works
Bird Detection
Each bird species has a unique acoustic signature. Spectrograms visualize these sounds, revealing distinct patterns that our ML model uses to identify birds.
Black Woodpecker · Palokärki
Common Chaffinch · Peippo
Common Crane · Kurki
System Architecture
Audio Recording
Remote Raspberry Pi captures short audio clips every 20 minutes
Edge Device
Audio stored on the Remote Pi; offline-first design for unreliable connectivity
Central Storage
Processing Pi receives synced recordings and stores them for processing
Audio Classifier
Deep learning model identifies bird species from audio
PostgreSQL
Bird detections stored to PostgreSQL
Web Interface
This page showcasing real-time data and insights
Audio Recording
Remote Raspberry Pi captures short audio clips every 20 minutes
Edge Device
Audio stored on the Remote Pi; offline-first design for unreliable connectivity
Central Storage
Processing Pi receives synced recordings and stores them for processing
Audio Classifier
Deep learning model identifies bird species from audio
PostgreSQL
Bird detections stored to PostgreSQL
Web Interface
This page showcasing real-time data and insights
Design Challenges
Unreliable Internet
Remote cottage has intermittent connectivity. System designed to work offline and sync in batches when connection is available.
Edge Processing
Recording happens on-site with minimal processing. Heavy ML inference runs on more accessible and powerful hardware.
Data Management
Thousands of recordings accumulated over time. Efficient storage, retrieval, and cleanup strategies are essential.
Built with