" Several Deep learning models are implemented to classify persian modal music into seven high correlated categories.
Model folder contains following models :
1 _ Bidirectional LSTM + Bidirectional GRU (inspired by Autoencoder architecture)
2 _ CNN1D / CNN2D (inspired by Autoencoder architecture)
3 _ Residual CNN (inspired by Autoencoder architecture) (1D & 2D)
4 _ CNN + Bidirectional LSTM/GRU (inspired by Autoencoder architecture)
5 _ Residual Bidirectional LSTM following by Bidirectional GRU (inspired by Autoencoder architecture)
Preprocessing folder contains preprocessing Audio files based on following features:
1 _ Chroma_CENS
2 _ MFCC
3 _ Mel Spectrogram