woheller69/whoBIRD

What it solves

whoBIRD is an Android application designed to identify bird species by their vocalizations in real time. It allows birdwatchers to identify birds from a database of over 6,000 species worldwide without needing an internet connection.

How it works

The app uses the BirdNET project's machine learning algorithms and TFLite library to analyze audio input from the device's microphone. It processes audio locally on the device to provide immediate identification. Users can customize the detection process using a high-pass filter to reduce background noise and adjust probability thresholds to control the accuracy of detections.

Who it’s for

It is intended for birding enthusiasts and nature observers who want to identify birds in the field, including in remote areas where internet access is offline.

Highlights

• Offline Identification: Performs all analysis on-device, requiring no internet connection for real-time detection.
• Extensive Database: Supports identification of over 6,000 bird species. -n- Location-Aware Filtering: Uses a meta model to check if a bird species is likely to be present at a specific location and time, with a weighted version to better detect migratory birds.
• Customizable Audio: Supports various audio sources (including USB microphones) and high-pass filtering to minimize low-frequency noise.
• Observation Tracking: Maintains a database of detections that can be backed up, shared, or deleted.
• Audio Recording: Ability to save .wav files for each detection on Android 12+ devices.