app.py

import streamlit as st
import numpy as np
from PIL import Image
from tensorflow.keras.models import load_model
import joblib
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.applications.inception_v3 import preprocess_input
import cv2
from BackPropogation import BackPropogation
from Perceptron import  Perceptron
from sklearn.linear_model import Perceptron
import tensorflow as tf
import joblib
import pickle
from numpy import argmax


# Load the tokenizer using pickle
with open(r'tokeniser.pkl', 'rb') as handle:
    loaded_tokenizer = pickle.load(handle)

# Load saved models
image_model = load_model('tumor_detection_model.h5')
dnn_model = load_model('imdb_model.h5')
loaded_model = tf.keras.models.load_model('sms_spam_detection_dnnmodel.h5')  
perceptron_model = joblib.load('perceptron_model.joblib')
backprop_model = joblib.load('backprop_model.pkl')


# Streamlit app     
st.title("Classification")

# Sidebar
task = st.sidebar.selectbox("Select Task", ["Tumor Detection", "Sentiment Classification"])

def preprocess_text(text):
    tokenizer = Tokenizer()
    tokenizer.fit_on_texts([text])
    sequences = tokenizer.texts_to_sequences([text])
    preprocessed_text = pad_sequences(sequences, maxlen=4)

    return preprocessed_text



def predict_dnn(preprocessed_text):
    preprocessed_text = preprocessed_text.reshape((1, 4))  # Adjust the shape according to your model's input shape

    prediction = dnn_model.predict(preprocessed_text)
    st.write("DNN Prediction:", prediction)
    
    
        
def predict_rnn(input_text):
    # Process input text similarly to training data
    encoded_input = loaded_tokenizer.texts_to_sequences([input_text])
    padded_input = tf.keras.preprocessing.sequence.pad_sequences(encoded_input, maxlen=10, padding='post')
    prediction = loaded_model.predict(padded_input)
    if prediction > 0.5:
        return "spam"
    else:
        return "ham"


def predict_custom_perceptron(preprocessed_text):
    perceptron = CustomPerceptron(epochs=10)  # Using the custom Perceptron
    prediction = perceptron.predict(preprocessed_text)
    st.write("Custom Perceptron Prediction:", prediction)
    
def predict_sklearn_perceptron(preprocessed_text):
    perceptron = SklearnPerceptron()  # Using the sklearn Perceptron
    prediction = perceptron.predict(preprocessed_text)
    st.write("Sklearn Perceptron Prediction:", prediction)
    
def predict_backpropagation(preprocessed_text):
    prediction = backprop_model.predict(preprocessed_text)
    st.write("Backpropagation Prediction:", prediction)

# make a prediction for CNN
def preprocess_image(image):
    image = image.resize((299, 299))
    image_array = np.array(image)
    preprocessed_image = preprocess_input(image_array)

    return preprocessed_image


def make_prediction_cnn(image, image_model):
    img = image.resize((128, 128))
    img_array = np.array(img)
    img_array = img_array.reshape((1, img_array.shape[0], img_array.shape[1], img_array.shape[2]))

    preprocessed_image = preprocess_input(img_array)
    prediction = image_model.predict(preprocessed_image)

    if prediction > 0.5:
        st.write("Tumor Detected")
    else:
        st.write("No Tumor")

if task == "Sentiment Classification":
    st.subheader("Choose Model")
    model_choice = st.radio("Select Model", ["DNN", "RNN", "Perceptron", "Backpropagation"])

    st.subheader("Text Input")
    text_input = st.text_area("Enter Text")

    if st.button("Predict"):
        # Preprocess the text
        preprocessed_text = preprocess_text(text_input)
        if model_choice == "DNN":
            predict_dnn(preprocessed_text)
        elif model_choice == "RNN":
            if text_input:
                prediction_result = predict_rnn(text_input)
                st.write(f"The message is classified as: {prediction_result}")
            else:
                st.write("Please enter some text for prediction")
        elif model_choice == "Custom Perceptron":
            predict_custom_perceptron(preprocessed_text)
        elif model_choice == "Sklearn Perceptron":
            predict_sklearn_perceptron(preprocessed_text)
        elif model_choice == "Backpropagation":
            predict_backpropagation(preprocessed_text)

else:
    st.subheader("Choose Model")
    model_choice = st.radio("Select Model", ["CNN"])

    st.subheader("Image Input")
    image_input = st.file_uploader("Choose an image...", type="jpg")

    if image_input is not None:
        image = Image.open(image_input)
        st.image(image, caption="Uploaded Image.", use_column_width=True)

        # Preprocess the image
        preprocessed_image = preprocess_image(image)

        if st.button("Predict"):
            if model_choice == "CNN":
                make_prediction_cnn(image, image_model)