Bird Identification with Neural Networks (Deep Learning)

kHz

4

8

mm:ss.fff

mm:ss.fff

00:01

00:02

00:03

00:04

00:05

00:06

00:07

00:08

00:09

  

Loading MP3. Please Wait ...

              

kHz

5

10

15

Informationen

On this page you can upload your own recordings (wav/mp3/m4a) of songs or calls of European birds. As a result of the analysis you will get a list of suggestions sorted with decreasing probability to determine the recorded species.

The method is based on training deep neural networks (deep learning). It was successfully implemented and evaluated in the LifeCLEF Bird Identification Task 2018, where it provided the best system to identify Central and South American birds.

The identification system for European bird species implemented on this page, mainly uses training material from the Animal Sound Archive Berlin, the Reference System of Bioacoustic Data and the community-based platform Xeno-Canto.

Publications

1. Lasseck M (2018) Audio-based Bird Species Identification with Deep Convolutional Neural Networks. In: Working notes of CLEF 2018


2. Lasseck M (2015) Towards Automatic Large-Scale Identification of Birds in Audio Recordings. In: Lecture Notes in Computer Science Vol.9283: pp 364-375