GUI tool for annotating sound events manually from a spectrogram and audio playback. Built by Gabriel Cano during a summer fellowship with the San Diego Zoo Conservation Technology Lab.
The application is meant to use WAV files as input and works best for those created by AudioMoth devices.
The application has been tested the most on macOS (Mojave and Catalina) but has also been used on Ubuntu 18.04.
Clone the repo. For help check these instructions.
Create a virtual environment and activate it. For help check these instructions.
Install pip if needed. For help check these instructions.
Download dependencies by running:
pip install -r requirements.txt
On Ubuntu: you may have to install some additional Debian packages (e.g. sudo apt install libasound2-dev). These include (but may not be limited to):
- libasound2-dev
To run program:
python3 main.py
File -> New Session
This opens a dialog box where you can type or choose the:
-
data path (the path of the folder where the WAV files are stored)
-
save path (file for csv annotation record (path is optional but filename is not optional (at time of this writing filename absence results in crash). File does not need to already exist.)
The "Minimum Duration" value you supply will force selections to be at least that many seconds long.
File -> Load Session
This opens a file browser where you need to select the configuration file (.yaml) saved when ending a previous session. It will open the session to the WAV file you left off on and it will continue saving to the same csv output file.
Drag the mouse over a part of the spectrogram. Press the play button or Action -> Play to listen to the selection. Press the save button or Action -> Save to save an annotation with the specified tag in the textfield above the buttons.
The annotation has a filepath (fp) which is the full filepath to where the audio file resides. Next is the tag which is specified by the user in the textfield. The next two entries are the start and end index of the selection in the original audio data. Finally the sampling rate (sr) is specified in case the data is loaded as a spectrogram with a different sampling rate than the original audio file.
Find the information you would like to edit in the table. Click on an entry to edit it. The original csv file is changed to match when you click elsewhere. Currently there is no way to delete a row but you can erase the entry.
Use the previous and next button or Action -> Previous and Action -> Next.
Saving a session is for convenience. It does not affect whether the csv file is saved or not because the annotations are always saved the moment they are created.
Python -> Quit Python
This brings up a dialog box where you are given the option to save the current session. If you choose yes you provide a filename and choose where to save it.
Warning: If you press the "x" button the application will close without saving your session (however your annotations are saved the moment you create them).
Create a session normally and use a dummy save file.
Simply start a new session again and instead of creating a new save file select the one you were using before (for some reason the dialog box will disappear behind the application, just move the application out of the way and press ok on the box). By default, the spectrogram shown will be the first. You can use Actions -> Goto to skip ahead.