/genre-classification-gtzan

Music Genre Classification and Recommendation System using librosa+keras.

Primary LanguageJupyter Notebook

Music Genre Classification

In this python notebook we explore the flaws of the GTZAN dataset and build several classification models. We use librosa to extract spectrograms from audio samples and keras to build and train the models. We discuss how CNN kernels might not be well fit for spectrograms and also provide a simple recommendation system based on cosine similarity and use it to find replicas in the dataset.