pakoromilas / clients_api

Python clients for inferencing magcil API

clients_api

Python clients for inferencing MagCIL Deep Audio API.

Deep Audio API provides access to a set of models that analyze audio signals in terms of:

• General auditory analysis (discriminate between music, speech, other sounds and silence)
• Musical classification (musical genres, moods and styles)
• Speaker characteristics (gender, speaking style etc)
• Environmental sound analysis (recognize quality of "soundscape")

Access to the API is provided through a simple Python client that sends audio data via an GRPC connection. Audio predictions are returned in a simple json format.

0. Install requiremets

virtualenv venv
source venv/bin/activate
pip3 install -r requirements.txt

You may also need to install portaudio and related dependencies. For ubuntu:

sudo apt-get install libasound-dev portaudio19-dev libportaudio2 libportaudiocpp0

1. Client

In order to request predictions for specific audio files use the following command:

python3 client.py -m deployed_model -i <audio_input> -u url -t token --username user

Where:

• deployed_model: is the name of the deployed ensemble model in server which encapsulates both preprocessing and pytorch models (this argument could be list of models). Select between ensemble_dali_4_class, ensemble_dali_emotions or both.
• audio_input: is the path of the input wav files to be tested.
• url: is the 'url:port' of the grpc server, e.g. localhost:50051
• token: is the token to be used for authentication
• user: is the email to be used for authentication

2. Real-time Client

In order to request predictions for real-time recorded audios, use the following commad:

python3 client-real-time.py -m deployed_model -u url -t token --username user

Where:

• deployed_model: is the name of the deployed ensemble model in server which encapsulates both preprocessing and pytorch models (this argument could be list of models). Select between 'ensemble_dali_4_class', 'ensemble_dali_emotions' or both.
• url: is the 'url:port' of the grpc server, e.g. localhost:50051
• token: is the token to be used for authentication
• user: is the email to be used for authentication

3. Youtube Client

In order to request prediction for an audio downloaded from a specific youtube url, use the following command:

python3 client-youtube.py -m deployed_model -i youtube_url -t token -u url --username user

Where:

• deployed_model: is the name of the deployed ensemble model in server which encapsulates both preprocessing and pytorch models (this argument could be list of models). Select between ensemble_dali_4_class, ensemble_dali_emotions or both.
• youtube_url : is the url of the youtube video to be used as audio input.
• url: is the 'url:port' of the grpc server, e.g. localhost:50051
• token: is the token to be used for authentication
• user: is the email to be used for authentication