app.py

import streamlit as st
import os,sys
import IPython
import librosa
import librosa.display
import matplotlib.pyplot as plt
import numpy as np
import pickle
#from tensorflow import keras
#from keras.models import load_model
import joblib



def wave_plot(data,sampling_rate):
    fig, ax = plt.subplots(1, 1)
    fig.set_size_inches(30.5, 18.5)
    ax.set_facecolor("black")
    fig.set_facecolor("black")
    plt.ylabel('Amplitude')
    plt.title("WAVEFORM",fontweight="bold")
    librosa.display.waveshow(data, sampling_rate,x_axis='s')
    st.pyplot(fig,use_container_width=True)
    return data

def prediction(data,sampling_rate):
    dict_values = {0:"neutral",1:"calm",2:"happy",3:"sad",4:"angry",5:"fear",6:"disgust",7:"pleasant surprise"}
    
    
    
    path = "CnnModel.h5"
    cnn_model = pickle.load(path)
    arr=[] 
    mfccs = np.mean(librosa.feature.mfcc(y=data, sr=sampling_rate, n_mfcc=40).T,axis=0) 
    arr.append(mfccs)
    X_test = np.expand_dims(arr, axis=2)
    predict = cnn_model.predict(X_test)
    predictions=np.argmax(predict,axis=1)
    detected_emotion = [dict_values[val] for val in predictions]
    original_title = '<p style="font-family:Courier; color:DarkSeaGreen; font-size: 18px;"><b>EMOTION DETECTED</b></p>'
    st.markdown(original_title, unsafe_allow_html=True)
    st.subheader(detected_emotion[0].upper())

   
    

def main():
    st.title("Emotion Classifier App")    
    menu = ["CNN Model","MLP Model","About"]
    choice = st.sidebar.selectbox("Menu",menu)
    col1, col2 = st.columns(2)
    if choice == "CNN Model":
        
        with col1:
            original_title = '<p style="font-family:Courier; color:DarkSeaGreen; font-size: 20px;"><b>CHOOSE AUDIO FILE</b></p>'
            st.markdown(original_title, unsafe_allow_html=True)
            audio_file = st.file_uploader("", type=['WAV', 'MP3', 'ogg'])
            if audio_file is not None:
                with col1:
                    data, sampling_rate = librosa.load(audio_file)
                    original_title = '<p style="font-family:Courier; color:DarkSeaGreen; font-size: 18px;"><b>WAVE FORM</b> </p>'
                    st.markdown(original_title, unsafe_allow_html=True)
                    data = wave_plot(data,sampling_rate)

                with col2 :
                    original_title = '<p style="font-family:Courier; color:DarkSeaGreen; font-size: 18px;"><b>PLAY AUDIO</b> </p>'
                    st.markdown(original_title, unsafe_allow_html=True)
                    st.write('\n')
                    st.write("\n")
                    st.write("\n")
                    st.audio(audio_file, format='audio/WAV', start_time=0)
                    st.write("\n")
                    st.write("\n")
                    st.write("\n")
                    st.write("\n")
                    st.write("\n")
                    prediction(data,sampling_rate) 
            
    else:
        
            original_title = '<p style="font-family:Courier; color:Teal; font-size: 18px;"><b>Speech Emotion Recognition (SER) can be defined as extraction of the emotional state of the speaker from his or her speech signal.We will be using different models to detect emotions for the audio input provided by the user.</b></p>'
            st.markdown(original_title, unsafe_allow_html=True) 
            original_title = '<p style="font-family:Courier; color:white;font-size: 18px;"></br><b>User will have two options for trying out : MLP, CNN<b><p>'
            st.markdown(original_title, unsafe_allow_html=True)
            original_title = '<br></br>'
            st.markdown(original_title, unsafe_allow_html=True) 
            original_title = '<p style="font-family:Courier; color:Teal; font-size: 25px;"><b>CREATED BY</b></p>'
            st.markdown(original_title, unsafe_allow_html=True)  
            st.write("\n")
            original_title = '<p style="font-family:Courier; color:Teal; font-size: 20px;"><b>SAKSHI DHYANI</b></p>'
            st.markdown(original_title, unsafe_allow_html=True)       


if __name__ == '__main__':
    main()