MUSIC GENRE CLASSIFICATION

Aim:

The main aim of the project is to identify the genre of the music clips.This is done using GTZAN dataset. With the help of PyTorch , Convolutional neural network is made of 3 convolutional and 4 linear layers . ReLu activation function and SGD optimizer with learning rate =0.0005 is used .

Basic Flow:

For the input the spectograms were generated which were then passed to the convolutional neural network and with the help of genre as labels the classification was done.

Genres:

  • Blues
  • Classical
  • Country
  • Disco
  • Hiphop
  • Jazz
  • Metal
  • Pop
  • Reggae
  • Rock

Accuracy

Train set : 100 %

Test set : 65 %

Accuracy plot :

Loss plot :

By using pretrained (ResNet18 Architecture )

Accuracy

Train set : 100 %

Test set : 100 %

Accuracy plot :

Loss plot :