SpectroTune---Genre-Classifier

Introduction

Music genre classification is an essential aspect of music information retrieval (MIR) and has numerous applications in the music industry, ranging from content-based music recommendation to indexing and organizing music databases. With the rise in digital music streaming platforms, such as Spotify, the ability to classify songs into genres automatically becomes increasingly significant.

The Problem

The primary challenge in music genre classification is to extract meaningful features from audio tracks and employ them to categorize tracks into respective genres. The task is further complicated by the subtle distinctions between some music genres, overlaps in musical elements across genres, and the dynamic nature of music genres over time.

Data Available

There are 2 sources of data:

• GTZAN Dataset - Music Genre Classification
• Spotify 3-Second Mel Spectrograms

How To Use

• Create a folder called Data/genres_original with the unzipped GTZAN file
• Run audio_noise_adder_async.py to add noise to the .wav files and save it
• Run image_gen_async.py to use the noisy .wav files and generate spectrograms
• Run relevant script for different ML approachs to genre classification

References

• Creating a Music Genre Classifier using a Convolutional Neural Network
• Music Genre Classification using Deep Learning (Audio and Video)
• CS229 Final Report - Music Genre Classification
• Tzanetakis, G., & Cook, P. (2002). Musical genre classification of audio signals. IEEE Transactions on Speech and Audio Processing, 10(5), 293-302.
• Choi, K., Fazekas, G., & Sandler, M. (2017). Automatic tagging using deep convolutional neural networks. In "ISMIR" (pp. 805-811).
• Dieleman, S., & Schrauwen, B. (2014). End-to-end learning for music audio. In "ICASSP, IEEE" (pp. 6964-6968).