Skip to content
This sample includes simeple CNN classifier for music and audio-folder dataloader just like ImageFolder in torchvision.
Python
Branch: master
Clone or download
Latest commit fe93bd2 Oct 31, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
checkpoint 1 Jan 22, 2018
data change Sep 9, 2018
test initial commit Jan 22, 2018
train initial commit Jan 22, 2018
.DS_Store
.gitattributes Initial commit Jan 22, 2018
.gitignore Update .gitignore Oct 30, 2018
LICENSE Initial commit Jan 22, 2018
README.md Update README.md Jan 26, 2018
batchspliter.py initial commit Jan 22, 2018
dataloader.py
main.py initial commit Jan 22, 2018
model.py initial commit Jan 22, 2018
util.py initial commit Jan 22, 2018

README.md

Introduction

In this repository, I inplementd AudioFolder function with data-loader and a full convolutional classifier for classical music. The AudioFolder function was included in "dataloader.py". And "batchspliter.py" can transfer normal music files to short audio splits which matches the format as training dataset.

Regretfully, after splitting audio, you need to manually select training data and testing data and put them in order like "train/a", "train/b", "test/a", and "test/b".

The including classifier is just a sample for data-loader function, so it works really bad for music classification. If you want to build up a more effective classifier, please use LSTM-RNN, sampleRNN and so on.

You can’t perform that action at this time.