Classifying a song by genre

In this case study, we will be classifying a song into one of 10 possible genres. Imagine a scenario where we are tasked to automatically classify the genre of a song without manually listening to it. This way, we can potentially minimize operational overload as far as possible.

Getting ready

The strategy we'll adopt is as follows:

1. Download a dataset of various audio recordings and the genre they fit into.
2. Visualize and contrast a spectrogram of the audio signal for various genres.
3. Perform CNN operations on top of a spectrogram:
   • Note that we will be performing a CNN 1D operation on a spectrogram, as the concept of translation does not apply in the case of audio recordings

4. Extract features...