This project focuses on the application of neural networks to perform music genre classification on the well-known GTZAN dataset. The dataset consists of 1,000 thirty-second audio tracks evenly distributed across 10 genres.
We aim to explore the potential of machine learning techniques, particularly neural networks, for accurately classifying music by genre. We achieved an accuracy of 95.85%, outperforming all models we have seen on Kaggle. This project implements feature extraction and two different neural networks for classification, one without using any deep learning frameworks and another using Keras. We further improve the performance of the Keras-based network using model blending techniques.