Automated Source Classification of Marmoset Vocalizations

App

Docker Usage

Clone the repository:

git clone https://github.com/pedro-varela1/Automated-Source-Classification-of-Marmoset-Vocalizations.git

cd Automated-Source-Classification-of-Marmoset-Vocalizations

Build the container by replacing it with the tag, container name and version you want:

docker build -t <tag>/<container_name>:<version> .

Finally, run the app:

docker container run -p 5000:5000 <tag>/<container_name>:<version>

In your browser, go to http://127.0.0.1:5000/.

App UI

Data Entry

• Audio in WAV format containing calls.
• CSV file containing at least three columns:
  • label: Label indicating what type of call it is (only calls with the label p will be used);
  • onset_s: Start time of the call in seconds, relative to the audio;
  • offset_s: End time of the call in seconds, relative to the audio.

Data Output

The data output will be a zip file containing images of the calls from each possible line that served as input for the marmoset classification model, as well as a CSV file with the model predictions added to each line.