Music Genre Classification - PyTorch

Music Genre classification using Convolutional Neural Networks on Spectrograms in PyTorch

Image source

Using the repository

Clone the repository to the machine where you want to run the model.

git clone https://github.com/adityashrm21/music-classification-cnn-pytorch.git

It is preferable but not necessary to use a GPU.

Preparing the dataset

Go to the root directory of the cloned repository and run the following commands:

wget http://opihi.cs.uvic.ca/sound/genres.tar.gz
tar -xvzf genres.tar.gz

Running the model

You can train the model and do the inference on test set using:

python3 train.py --root_dir "." --lr 1e-3 --momentum 0.9 --epochs 50

Here is a complete list of arguments:

usage: train.py [-h] --root_dir ROOT_DIR [--epochs EPOCHS]
                [--batch_size BATCH_SIZE] [--lr LR] [--momentum MOMENTUM]
                [--weight_decay WEIGHT_DECAY]

optional arguments:
  -h, --help            show this help message and exit
  --root_dir ROOT_DIR   root directory for the dataset
  --epochs EPOCHS       num of training epochs
  --batch_size BATCH_SIZE
                        training batch size
  --lr LR               learning rate
  --momentum MOMENTUM   momentum for SGD
  --weight_decay WEIGHT_DECAY
                        weight decay for L2 penalty

Results

Current test accuracy:

• without a pretrained model and 100 epochs: 75%
• with a pretrained model and 75 epochs: 60%