setup.py

#This .py file contains all of the imports and global variable definitions that are used throughout the project.
import librosa
import numpy as np
from functools import partial
import pandas as pd
import matplotlib.pyplot as plt
import io
import cv2
import matplotlib.colors as colors
import matplotlib.pyplot as plt
import mpl_toolkits.mplot3d as a3
import sys
import os
from pedalboard import Pedalboard, Distortion, Bitcrush, Resample
from pedalboard.io import AudioFile
import pyroomacoustics as pra
from multiprocessing import Pool
import random
import IPython
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay
from sklearn.preprocessing import OneHotEncoder
from sklearn.metrics import precision_score, recall_score, f1_score
import scipy.io.wavfile

sr = 48000 #Samplerate at which we will be working
udb = (-120, 10) #Decibel range for the spectrogram generation
uflim = (20, 20000) #Frequency range for the spectrogram generation
plot_directory = "/Users/ivanstankov/Documents/Study/thesis_code/imgs/" #Directory where the plots will be saved

def load_audio(path, sr):
    """
    This function is used to read audio file into array.
    It also allows user to read audio data with target samplerate and convert stereo signal to mono.
    """
    # We use a popular audio tool called Librosa.
    data, _ = librosa.load(path, sr=sr, mono=True)
    data = librosa.util.normalize(data) 
    return data


def save_audio(path, data, sr):
    """
    This function is used to save audio data into audio file.
    """
    # Scipy is used instead of librosa because librosa has some issues with saving audio files.
    scipy.io.wavfile.write(path, sr, data)