Exploring Neural Nets in Audio context to familiarize with DSP concepts by working through the multi-class genre classification problem using 10 genres/labels.
This repo includes code that
-
explores fundamentals of neurons and neural networks by implementing MLP from scratch
-
uses Tensorflow to replicate the above
-
builds on basics of DSP/audio processing relevant for prepping datasets for deep learning,
-
solves genre classification using various Neural Net architectures (MLP/ CNN/ RNN/ RNN + LSTM)
-
digs into nuances involved in improving model performance and keeping the models interpretable
The code is indexed from 0. Some of it is standalone, while some build on the previous index.
Train/Test Dataset: https://www.kaggle.com/andradaolteanu/gtzan-dataset-music-genre-classification