riyageorge/Multitasking_App

1import streamlit as st2from PIL import Image3import tensorflow as tf4import numpy as np5from tensorflow.keras.datasets import imdb6from tensorflow.keras.preprocessing import sequence7from numpy import argmax8import pickle9 10 11 12# Load your CNN tumor classification model13cnn_model = tf.keras.models.load_model('cnn_tumor_model.h5')14 15# Function to perform image classification using CNN16def classify_image(img, cnn_model):17    img = Image.open(img)18    img = img.resize((128, 128))19    img = np.array(img)20    input_img = np.expand_dims(img, axis=0)21    res = cnn_model.predict(input_img)22    if res > 0.5:23        return "Tumor Detected"24    else:25        return "No Tumor"26 27 28# Load your DNN SMS spam detection model29dnn_smsspam_model = tf.keras.models.load_model('dnn_smsspam_model.h5')30# Load the saved tokenizer31with open('dnn_smsspam_tokenizer.pickle', 'rb') as handle:32    dnn_smsspam_tokenizer = pickle.load(handle)33 34def dnn_predict_message(input_text):35    max_length=2036    # Process input text similarly to training data37    encoded_input = dnn_smsspam_tokenizer.texts_to_sequences([input_text])38    padded_input = tf.keras.preprocessing.sequence.pad_sequences(encoded_input, maxlen=max_length, padding='post')39    # Get the probabilities of being classified as "Spam" for each input40    predictions = dnn_smsspam_model.predict(padded_input)41    # Define a threshold (e.g., 0.5) for classification42    threshold = 0.543    # Make the predictions based on the threshold for each input44    for prediction in predictions:45        if prediction > threshold:46            return "Spam"47        else:48            return "Not spam"49 50 51# Load your RNN SMS spam detection model52rnn_smsspam_model = tf.keras.models.load_model('rnn_smsspam_model.h5')53# Load the saved tokenizer54with open('rnn_smsspam_tokenizer.pickle', 'rb') as handle:55    rnn_smsspam_tokenizer = pickle.load(handle)56 57def rnn_predict_message(input_text):58    max_length=2059    # Process input text similarly to training data60    encoded_input = rnn_smsspam_tokenizer.texts_to_sequences([input_text])61    padded_input = tf.keras.preprocessing.sequence.pad_sequences(encoded_input, maxlen=max_length, padding='post')62    # Get the probabilities of being classified as "Spam" for each input63    predictions = rnn_smsspam_model.predict(padded_input)64    # Define a threshold (e.g., 0.5) for classification65    threshold = 0.566    # Make the predictions based on the threshold for each input67    for prediction in predictions:68        if prediction > threshold:69            return "Spam"70        else:71            return "Not spam"72 73            74# Load the saved LSTM model75lstm_smsspam_model=tf.keras.models.load_model('lstm_smsspam_model.h5')76# Load the saved tokenizer77with open('lstm_smsspam_tokenizer.pickle', 'rb') as handle:78    lstm_smsspam_tokenizer = pickle.load(handle)79 80def lstm_predict_message(message):81    maxlen=5082    sequence = lstm_smsspam_tokenizer.texts_to_sequences([message])83    sequence = tf.keras.preprocessing.sequence.pad_sequences(sequence, padding='post', maxlen=maxlen)84    prediction = lstm_smsspam_model.predict(sequence)[0, 0]85    if prediction > 0.5:86        return 'Spam'87    else:88        return 'Not spam'89 90 91# Load the saved model92gru_movie_model = tf.keras.models.load_model('gru_movie_model.h5')93with open('tokenizer_movie_gru.pickle', 'rb') as handle:94    lstm_movie_tokeniser = pickle.load(handle)95 96# Function to predict sentiment for a given review97def gru_predict_movie_sentiment(review):98    maxlen = 10099    sequence = lstm_movie_tokeniser.texts_to_sequences([review])100    sequence = tf.keras.preprocessing.sequence.pad_sequences(sequence, padding='post', maxlen=maxlen)101    prediction = gru_movie_model.predict(sequence)102    if prediction > 0.5:103        return "Positive"104    else:105        return "Negative"106 107 108with open('perceptron_movie_model.pkl', 'rb') as file:109    perceptron_movie_model = pickle.load(file)110 111def predict_movie_sentiment_perceptron(review):112    max_review_length = 500113    top_words = 5000114    word_index = imdb.get_word_index()115    review = review.lower().split()116    review = [word_index[word] if (word in word_index and word_index[word] < top_words) else 0 for word in review]117    review_bin = np.where(np.array(review) > 0, 1, 0)118    # Padding or truncating the review to match the perceptron's input size119    review_bin_padded = np.pad(review_bin, (0, max_review_length - len(review_bin)), 'constant')120    prediction = perceptron_movie_model.predict([review_bin_padded])121    if prediction[0] == 1:122        return "Positive"123    else:124        return "Negative"125 126        127# Load the saved instance of the Perceptron class128with open('backprop_movie_model.pkl', 'rb') as file:129    backprop_movie_model = pickle.load(file)130 131def predict_movie_sentiment_backprop(review):132    max_review_length = 500133    top_words = 5000134    word_index = imdb.get_word_index()135    review = review.lower().split()136    review = [word_index[word] if (word in word_index and word_index[word] < top_words) else 0 for word in review]137    review_bin = np.where(np.array(review) > 0, 1, 0)138    # Padding or truncating the review to match the perceptron's input size139    review_bin_padded = np.pad(review_bin, (0, max_review_length - len(review_bin)), 'constant')140    prediction = backprop_movie_model.predict([review_bin_padded])141    if prediction[0] == 1:142        return "Positive"143    else:144        return "Negative"145 146 147 148# Main function for Streamlit app149def main():    150    st.title("Multitasking App")    151        152    # Sidebar dropdown for selecting tasks153    task = st.sidebar.radio("Select Task", (["Tumor Detection", "Sentiment Classification"]))154 155    # Depending on the selected task, provide model options156    if task == "Tumor Detection":157        model = st.sidebar.radio("Select Model", (["CNN"]))158        159        if model == "CNN":160            st.subheader("Tumor Detection")161            uploaded_file = st.file_uploader("Upload an image to check for tumor...", type=["jpg", "png", "jpeg"])162 163            if uploaded_file is not None:164                # Display the image165                image_display = Image.open(uploaded_file)166                st.image(image_display, caption="Uploaded Image", use_column_width=True)167                168                if st.button("Detect Tumor"):169                    # Call the tumor detection function170                    result = classify_image(uploaded_file, cnn_model)171                    st.write("Tumor Detection Result:", result)172        173        174 175    elif task == "Sentiment Classification":176        model = st.sidebar.radio("Select Model", (["DNN", "RNN", "LSTM", "GRU", "Perceptron", "Backpropagation"]))177        178        179 180        if model == "DNN":181            st.subheader("SMS Spam Detection")182            user_input = st.text_area("Enter a message to classify as 'Spam' or 'Not spam': ")183                184            if st.button("Predict"):185                if user_input:186                    prediction_result = dnn_predict_message(user_input)187                    st.write(f"The message is classified as: {prediction_result}")188                else:189                    st.write("Please enter some text for prediction")190 191        elif model == "RNN":192            st.subheader("SMS Spam Detection")193            user_input = st.text_area("Enter a message to classify as 'Spam' or 'Not spam': ")194                195            if st.button("Predict"):196                if user_input:197                    prediction_result = rnn_predict_message(user_input)198                    st.write(f"The message is classified as: {prediction_result}")199                else:200                    st.write("Please enter some text for prediction")201 202        elif model == "LSTM":203            st.subheader("SMS Spam Detection")204            user_input = st.text_area("Enter a message to classify as 'Spam' or 'Not spam': ")205                206            if st.button("Predict"):207                if user_input:208                    prediction_result = lstm_predict_message(user_input)209                    st.write(f"The message is classified as: {prediction_result}")210                else:211                    st.write("Please enter some text for prediction")212 213        elif model == "GRU":214            st.subheader("Movie Sentiment Analysis")215            user_review = st.text_area("Enter a movie review: ")216            217            if st.button("Analyze Sentiment"):218                if user_review:219                    sentiment_result = gru_predict_movie_sentiment(user_review)220                    st.write(f"The sentiment of the review is: {sentiment_result}")221                else:222                    st.write("Please enter a movie review for sentiment analysis")223 224        elif model == "Perceptron":225            st.subheader("Movie Sentiment Analysis")226            user_review = st.text_area("Enter a movie review: ")227            228            if st.button("Analyze Sentiment"):229                if user_review:230                    sentiment_result = predict_movie_sentiment_perceptron(user_review)231                    st.write(f"The sentiment of the review is: {sentiment_result}")232                else:233                    st.write("Please enter a movie review for sentiment analysis")234                    235        elif model == "Backpropagation":236            st.subheader("Movie Sentiment Analysis")237            user_review = st.text_area("Enter a movie review: ")238            239            if st.button("Analyze Sentiment"):240                if user_review:241                    sentiment_result = predict_movie_sentiment_backprop(user_review)242                    st.write(f"The sentiment of the review is: {sentiment_result}")243                else:244                    st.write("Please enter a movie review for sentiment analysis")245 246                247if __name__ == "__main__":248    main()249