get_spectrogram.py

import os
import pandas as pd
import librosa
import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
import ipdb
import argparse
from p_tqdm import p_map
import warnings
warnings.filterwarnings('ignore')

parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter, description='TODO')
parser.add_argument('-f','--file', type=str,required=True,help = 'Name of the file that contain the recording to print')
parser.add_argument('-p','--path_to_data', type=str, help = 'Path of the folder that contain the recodings',required=True)
parser.add_argument('-d','--direction', type=str, help = 'Directory to wich spectrogram will be stored',required=True)
parser.add_argument('-m','--mode', type=str,choose={'unique','multiple'} help = 'Direction of the saved spectrogram',required=True)
args = parser.parse_args()

path_to_data = args.path_to_data
direction = args.direction
folder = 'Spectrogram/'
DURATION = 5
OVERLAP = 2
NB_IMG_PER_REC = 1

count = 0

df = pd.read_csv(str(args.file),low_memory=False)
df['Path'] = ''
for w in range (len(df)):
    df.loc[w,'Path']=str(args.path_to_data+ df.code_unique[w]+"_split_" + str(df.chunk_initial_time[w]) + "_" + str(df.chunk_final_time[w]))


def process(x):
#for i in tqdm(df.groupby('Path')):  #la colonne dir correspond au path
   count, (i) = x

   for j in range (NB_IMG_PER_REC): #30*8 secondes - 30*2 secondes (overlap) = 180 secondes affichées sur 30 images : n'affiche que les 3 premières minutes d'un enregistrement

        filename = str(i[0])

        if args.mode == 'multiple' :
            if count ==0:
                duration = DURATION
                offset = 0 
            else :
                duration = DURATION
                offset = offset + (DURATION - OVERLAP) 
        if args.mode == 'unique':
            offset = 0

        try:
            y, sr = librosa.load(filename, offset = offset, duration = DURATION)
        except Exception:
            continue
        window_size = 1024
        window = np.hanning(window_size)
        stft  = librosa.core.spectrum.stft(y, n_fft=window_size, hop_length=512, window=window)
        out = 2 * np.abs(stft) / np.sum(window)

        plt.close()
        plt.figure()

        vmin = np.flipud(np.log10(np.abs(stft))).mean()
        vmax = np.flipud(np.log10(np.abs(stft))).max()

        #a = plt.specgram(y, Fs=sr, NFFT=window_size, noverlap=512, cmap='jet', vmin=vmin*10, vmax=vmax*10)
        plt.imshow(np.flipud(np.log10(np.abs(stft))),aspect = "auto",interpolation = None,cmap = 'jet',vmin = vmin,vmax = vmax)

        plt.subplots_adjust(top=1, bottom=0, left=0, right=1)

        name = str(i[0].replace('/','_').split('.')[0]+'_'+str(count))
        try :

            plt.savefig(str(direction+folder+name+'.jpg'))

        except FileNotFoundError:

            os.mkdir(str(direction+folder))
            plt.savefig(str(direction+folder+name+'.jpg'))

        #count+=1

p_map(process, enumerate(df.groupby('Path')), num_cpus=10) #la colonne dir correspond au path