Python Flask application which generates a guitar tab display based on the submitted audio recording.
The recording is split into segments based on the loudness (amplitude) level changes.
A segment is considered a piece of the signal in which the amplitude rises above and falls below the user's input db level.
Each segment is visually represented in the form of a Mel Spectrogram image.
The images are used as inputs based on which a trained CNN model gives predictions of the musical notes in the images.
Once an output array of notes is returned as a response to the submitted request, a responsive guitar tablature is represented dynamically to the user.
Output tablature example
A loudness chart graph is also displayed to the user so that detection errors can be avoided by setting a different loudness threshold and re-submitting the request.
Output chart example
Install the following packages:
pip install opencv-python librosa flask tensorflow matplotlib -r requirements