app.py

import numpy as np 
import librosa
import math
from tensorflow import keras
import streamlit as st
import time
import pyautogui

def get_mfcc(audio_signal, num_mfcc=13, n_fft=2048, hop_length=512, num_segments=5):
    new_data = {
        "mfcc": []
    }

    SAMPLE_RATE = 22050
    signal,sample_rate = librosa.load(audio_signal,sr = SAMPLE_RATE)
    TRACK_DURATION = int(librosa.get_duration(signal)) # measured in seconds
    SAMPLES_PER_TRACK = SAMPLE_RATE * TRACK_DURATION
    samples_per_segment = int(SAMPLES_PER_TRACK / num_segments)
    num_mfcc_vectors_per_segment = math.ceil(samples_per_segment / hop_length)

    for d in range(num_segments):

        start = samples_per_segment * d
        finish = start + samples_per_segment

        mfcc = librosa.feature.mfcc(signal[start:finish], sample_rate, n_mfcc=num_mfcc, n_fft=n_fft, hop_length=hop_length)
        mfcc = mfcc.T

        if len(mfcc) == num_mfcc_vectors_per_segment:
            new_data["mfcc"].append(mfcc.tolist())

    return new_data["mfcc"]

def prediction(mfcc):

    cnn_model = keras.models.load_model('music-gen-classify-v1/')
    mfcc = np.array(mfcc)
    mfcc = mfcc[...,np.newaxis]
    prediction = cnn_model.predict(mfcc)
    return max(np.argmax(prediction,axis = 1))

def get_genre(prediction):

    pred = ''
    if prediction == 0:
        pred = 'Blues'
        
    elif prediction == 1:
        pred = 'Classical'
    
    elif prediction == 2:
        pred = 'Country'
    
    elif prediction == 3:
        pred = 'Disco'
    
    elif prediction == 4:
        pred = 'Hip Hop'
        
    elif prediction == 5:
        pred = 'Jazz'    
    
    elif prediction == 6:
        pred = 'Metal'
    
    elif prediction == 7:
        pred = 'Pop'
    
    elif prediction == 8:
        pred = 'Reggae'
    
    elif prediction == 9:
        pred = 'Rock'
    
    return pred


def main():

    st.set_page_config(layout='wide',page_title='Klasifikasi Genre Musik',page_icon='🎵')
    st.title('Prediksi Genre Musik')
    st.markdown('Website ini bisa memprediksi hingga 10 genre yaitu Blues, Classical, Country, Disco, Hiphop,Jazz, Metal, Pop, Reggae, dan Rock. Tentu prediksi genre ini tidak selalu tepat karena lagu itu multi genre, tapi selamat mencoba!')        
    selected_item = st.selectbox('Pilih Sample Musik Atau Upload Musikmu !',['Sample Musik','Upload'])

    if selected_item is not None:
        if selected_item == 'Upload':
            files = st.file_uploader('Format Musik ".wav" berdurasi maksimal 30sec', type='wav', accept_multiple_files=False)
            
            if files is not None:
                audio,sr = librosa.load(files,sr = 22050)
                duration = int(librosa.get_duration(audio))
                if 'file_uploaded' not in st.session_state:
                    st.session_state['file_uploaded'] = True
                
                    for percent_complete in range(100):
                        time.sleep(0.01)
                        bar.progress(i+1)
                        i = i+1       
                    pyautogui.hotkey('ctrl', 'F5')

                elif st.session_state['file_uploaded'] == True:
                    st.audio(files, format="audio/wav", start_time=0)              
        
        
        elif selected_item == 'Sample Musik':
            selected_file = st.selectbox("Pilih Sample", ['Blues','Jazz','Country','Classical','Hiphop','Metal','Pop','Reggae','Rock'])
            files = f'Data/upload/user/{selected_file}.wav'
            st.audio(files, format="audio/wav", start_time=0)
        submitted = st.button("Prediksi")
        
    if submitted:
        with st.spinner('Tunggu Sebentar'):
            signal = files
            mfcc_for_track = get_mfcc(signal)
    
            predict = prediction(mfcc_for_track)
            genre = get_genre(int(predict))
        st.success('Proses Prediksi Selesai!')
        st.markdown(f'Genre Musikmu adalah 🎵  : Musik **{genre}** ')

if __name__ == '__main__':
    main()