import streamlit as st
from transformers import pipeline
import matplotlib.pyplot as plt
from PIL import Image, ImageFilter
import numpy as np
from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2, preprocess_input, decode_predictions
from tensorflow.keras.preprocessing.image import img_to_array

# Chargement du modèle d'analyse de sentiments et de reconnaissance d'images
sentiment_pipeline = pipeline("sentiment-analysis")
image_model = MobileNetV2(weights='imagenet')

def plot_sentiment(sentiment):
    """ Dessine un graphique en camembert pour l'analyse de sentiments. """
    labels = ['Positive', 'Negative']
    sizes = [sentiment['score'], 1 - sentiment['score']]
    colors = ['#ff9999','#66b3ff']
    fig, ax = plt.subplots()
    ax.pie(sizes, colors=colors, labels=labels, autopct='%1.1f%%', startangle=90)
    ax.axis('equal')
    return fig

def classify_image(image):
    """ Classifie l'image et retourne les prédictions. """
    image = image.resize((224, 224))
    image = img_to_array(image)
    image = np.expand_dims(image, axis=0)
    image = preprocess_input(image)
    preds = image_model.predict(image)
    return decode_predictions(preds, top=3)[0]

def apply_filters(image):
    """ Applique des filtres à l'image et retourne l'image filtrée. """
    filter = st.sidebar.selectbox("Sélectionnez un filtre", ["Original", "Flou", "Contours", "Accentuation"])
    if filter == "Flou":
        return image.filter(ImageFilter.BLUR)
    elif filter == "Contours":
        return image.filter(ImageFilter.CONTOUR)
    elif filter == "Accentuation":
        return image.filter(ImageFilter.EDGE_ENHANCE)
    return image

# Configuration de l'application
st.title("Mon Application de Reconnaissance d'Images et d'Analyse de Sentiments")
st.sidebar.title("Options")

# Barre de navigation
options = st.sidebar.radio("Choisissez une fonctionnalité:", ['Reconnaissance d\'Images', 'Analyse de Sentiments'])

# Reconnaissance d'images
if options == 'Reconnaissance d\'Images':
    uploaded_file = st.file_uploader("Téléchargez une image...", type=["jpg", "png", "jpeg"])
    if uploaded_file is not None:
        image = Image.open(uploaded_file)
        filtered_image = apply_filters(image)
        st.image(filtered_image, caption='Image téléchargée', use_column_width=True)
        
        # Classification de l'image
        st.write("Classifying...")
        labels = classify_image(filtered_image)
        for label in labels:
            st.write(f"{label[1]} ({label[2]*100:.2f}%)")

        # Analyse de sentiments basée sur les étiquettes de l'image
        sentiments = [sentiment_pipeline(label[1])[0] for label in labels]
        for sentiment in sentiments:
            st.write(f"Sentiment pour '{sentiment['label']}' : {sentiment['score']:.2f}")
            fig = plot_sentiment(sentiment)
            st.pyplot(fig)

# Analyse de sentiments
elif options == 'Analyse de Sentiments':
    user_input = st.text_area("Entrez le texte à analyser", "Tapez ici...")
    if st.button("Analyser"):
        result = sentiment_pipeline(user_input)[0]
        st.write(f"Sentiment: {result['label']}, Score: {result['score']:.2f}")
        fig = plot_sentiment(result)
        st.pyplot(fig)