Welcome to DeepSS

DeepSS — short for DeepSongSegmenter — is a tool for annotating song in audio recordings. At the core of DeepSS is a deep neural network, implemented in Tensorflow. The network takes single- and multi-channel audio as an input and returns the probability of finding a particular song type for each audio sample. DeepSS can be used with a graphical user interface for loading audio data, annotating song manually, training a network, and generating annotations on audio. Alternatively, DeepSS can be used programmatically from the command line, in python notebooks, or in your own python code via the dss module.

If you use DeepSS, please cite:

Elsa Steinfath, Adrian Palacios, Julian Rottschäfer, Deniz Yuezak, Jan Clemens (2021). Fast and accurate annotation of acoustic signals with deep neural networks, bioRxiv,


Quick start tutorial

Annotate song, train a network, and predict on new samples.

Using the GUI

Comprehensive description of all GUI dialogs and options.

Use in python and from the terminal

Convert your own data, train and evaluate a network, predict on new samples in realtime.


Discover song types in annotated syllables.