Explainable Audio Models (ExAMod)

Code for the paper: Changhong Wang, Vincent Lostanlen, and Mathieu Lagrange. Explainable Audio Classification of Playing Techniques with Layer-wise Relevance Propoagation, submitted to IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023.

How to run

Get code

git clone https://github.com/changhongw/examod.git

Install dependencies

conda env create -f environment.yml
conda activate examod

Data

This work uses the Studio On Line (SOL) dataset (version 0.9HQ). Please contact the dataset creators for version 0.9HQ if you would like to reproduce all the results in our paper.

After downloading the dataset, the SOL-PMT subset and the meta data can be automatically generated by running

0_data_meta.ipynb

The data split and the corresponding file IDs used in our paper is shown in the SOL-0.9HQ-PMT_meta.csv file.

Carrier-modulation feature map extraction

python 1_preprocess_feature.py
(feature extraction takes around 0.5 h on one GPU)

Playing technique classification

Train and test network python 2_classification.py
(classification with the full scattering feature takes around 15 mins on one GPU)
Check classification result 3_classification_results.ipynb

Explanation maps

Local evidence map 4_local_relevance_maps.ipynb
Class-wise aggregation 5_classwise_aggregation.ipynb

Please feel free to open an issue if you find any bug.