Anaconda python: Install the anaconda python distribution (or miniconda). If conda is already installed on your system, make sure you have conda v4.8.4+. If not, update from an older version with conda update conda.

CUDA libraries for using the GPU: While DeepSS works well for annotating song using CPUs, GPUs will greatly improve annotation speed and are highly recommended for training. DeepSS uses Tensorflow as a deep-learning backend. To ensure that Tensorflow can utilize the GPU, the required CUDA libraries need to be installed. See the tensorflow docs for details.

Libsoundfile on linux: If you are on linux and want to load audio from a wide range of audio formats (other than wav), then you need to install libsndfile. The GUI uses the soundfile python package, which relies on libsndfile. libsndfile will be automatically installed on Windows and macOS. On Linux, the library needs to be installed manually with: sudo apt-get install libsndfile1. Again, this is only required if you need to load data from more exotic audio files.

Visual C++ runtime on windows: This is typically installed so only required if dss fails to load the native tensorflow runtime. Download the latest version from here.

Install DeepSS with or without the GUI

Create an anaconda environment called dss that contains all the required packages, including the GUI:

conda env create -f -n dss

If you do not need the graphical user interface (for instance, when training DeepSS on a server), install the non-GUI version:

conda env create -f -n dss


Don’t. It only causes problems in our experience. Best to install into a fresh environment than to update an existing environment. The brave can update using pip:

conda activate dss
pip install deepss --update  # DeepSS itself
pip install xarray_behave --update  # the GUI

Test the installation (Optional)

To quickly test the installation, run these commands in the terminal:

conda activate dss  # activate the conda environment
dss train --help  # test dss training
dss gui  # start the GUI

The second command will display the command line arguments for dss train. The last command, dss gui, will start the graphical user interface - this step will not work with the non-GUI install.

Make a clickable desktop icon (Optional)

To start the DeepSS GUI without having to use a terminal, create a clickable startup script on the desktop.

On macOS or linux, place a text file called (linux) or xb.command (macOS) with the following content on the desktop:

# /bin/bash
source $CONDA_PREFIX/etc/profile.d/
conda activate dss
dss gui

Make the files executable with chmod +x FILENAME, where FILENAME is (linux) or xb.command (macOS).

For windows, place a text file called xb.bat with the following content on the desktop:

CALL conda.bat activate dss
dss gui

Next steps

If all is working, you can now use DeepSS to annotate song. To get started, you will first need to train a network on your own data. For that you need annotated audio - either create new annotations using the GUI or convert existing annotations using python scripts.