Project folder used to train an audio genre classifier using a CNN
Navigating files:
- (audio_eda_feature_extraction) Import data and extract features/sample EDA
- (librosa) Explore Librosa to create spectrograms
- (create_spectrograms) Create mel spectrograms for both 3 and 10 second segments
- (split-data) Splitting data (spectrograms) into train, val, test
- Keras/PyTorch CNN models
CNN created with PyTorch performed way better than Tensorflow/Keras.