source code for the paper publised in IJCNN 2020 "Regression-based Music Emotion Prediction using Triplet Neural Networks"

natsort==7.0.1

tensorflow-GPU==1.13.1

keras==2.3.1

seaborn==0.10.0

scikit-learn==0.19.2

pydub==0.23.1

matplotlib==3.0.2

tqdm

The train the model, please refer to Training.ipynb. The model is trained on DEAM dataset, which can be downloaded here. The data processing steps can be found in DEAM/Data_Processing.ipynb.

To use your trained model to predict music emotion, please refer to Demonstration.ipynb.

In each of the region, we picked one song, and the predicted song can be found in /Demo/*

Low

UpperLow

Middle

LowerHigh

High