music-genres-classification

Abstract

The demo is to develop a deep learning model that will identify the genres from music.The model adopt Inception-ResnetV2 from slim.And the datasat used GTZAN Music Genre Dataset, which is a collection of 1000 songs in 10 genres, is the most widely used dataset.If you wanna train a model by yourself, download it from GTZAN dataset.

Below table shows the result on test set:

Accuracy	Value
Top-1	69.70%
Top-3	92.50%

Below picture shows the training-process on tensorboard:

Loss

Accuracy

Pre-Trained Testing

• Step1: download pre_trained model,and put it into models/ folder.
• Step2: test by executing the following command:

python test.py

Training

• Step1: download Dataset GTZAN,and put it into GTZAN/ folder.
• Step2: create the tfrecords by executing the following command:

usage: create_data_to_train.py [-h] [--data_path DATA_PATH]
                               [--train_path TRAIN_PATH]
                               [--test_path TEST_PATH] [--test_size TEST_SIZE]
                               [--aug AUG]

optional arguments:
  -h, --help            show this help message and exit
  --data_path DATA_PATH
                        data_sets path.
  --train_path TRAIN_PATH
                        train tfrecords save path.
  --test_path TEST_PATH
                        test tfrecords save path.
  --test_size TEST_SIZE
                        Proportion of test data that between [0, 1]
  --aug AUG             the size of data sets up to (arg) times as original. 1
                        means not augmentation.                        

• Step3: train the model by executing the following command:

usage: train.py [-h] [--train_data TRAIN_DATA] [--decay_rate DECAY_RATE]
                [--global_steps GLOBAL_STEPS] [--decay_steps DECAY_STEPS]
                [--learning_rate LEARNING_RATE] [--eval_step EVAL_STEP]
                [--batch_size BATCH_SIZE] [--model_path MODEL_PATH]
                [--summary_path SUMMARY_PATH] [--load_model LOAD_MODEL]

optional arguments:
  -h, --help            show this help message and exit
  --train_data TRAIN_DATA
                        train_data path.
  --decay_rate DECAY_RATE
                        learning rate decay rate.
  --global_steps GLOBAL_STEPS
                        global steps
  --decay_steps DECAY_STEPS
                        learning rate decay steps.
  --learning_rate LEARNING_RATE
                        learning rate.
  --eval_step EVAL_STEP
                        evaluation steps.
  --batch_size BATCH_SIZE
                        batch size.
  --model_path MODEL_PATH
                        tensorflow model path.
  --summary_path SUMMARY_PATH
                        tensorflow summary path.
  --load_model LOAD_MODEL
                        whether you wish to continue training.

Testing

Run the following command to simple test the test-dataset:

python test.py

References

[1] https://github.com/deepsound-project/genre-recognition