Example systems for Chapter 2, The Machine Learning Approach for Analysis of Sound Scenes and Events, Toni Heittola, Emre Cakir, and Tuomas Virtanen, In book "Computational Analysis of Sound Scenes and Events", Ed. Virtanen, T. and Plumbley, M. and Ellis, D., pp.13-40, 2018, Springer
Applications are based on multilayer perceptron (MLP) approach, and all of them are implementing similar overall system structure. Implementation is done in Python with Keras machine learning library.
single_label_classification.py
An example system to show how to tackle single-label classification problem in audio analysis. Acoustic scene classification is used as example application with TUT Acoustic Scenes 2017, development dataset. The setup is similar to DCASE2017 baseline system for Task 1. The dataset contains 10 second long audio excerpts from 15 different acoustic scene classes.
multi_label_classification.py
An example system to show how to tackle multi-label classification problem in audio analysis. Audio tagging is used as example application with [CHiME-Home, development & evaluation dataset](CHiME-Home, development & evaluation dataset).
sound_event_detection.py
An example system to show how to tackle detection problem in audio analysis. Sound event detection is used as example application with TUT Sound events 2017, development dataset.
- Clone repository from Github or download latest release.
- Install requirements with command:
pip install -r requirements.txt - Run the system:
python single_label_classification.py
System will download a benchmark dataset (stored under temp directory), train acoustic models based on it and evaluate system's performance.
Code released under the license from Tampere University of Technology. Code is free to use for experimental and non-commercial purposes.