Searching for similar audios has a wide range of application including finding similar songs, replacing curse words and detecting the speakers. In this tutorial, we will build an example of searching similar audios using the AudioSet dataset and the VGGish model.
Download the model
bash scripts/download_models.sh
Start the flow and index the data at ./toy-data
python app.py
With the Flow running as a http service, we can use the Jina swagger UI tool to query.
Open the browser at localhost:45678/docs
, send query via the Swagger UI,
{
"data": [
{
"uri": "toy-data/6pO06krKrf8_30000_airplane.mp3"
}
]
}