import streamlit as st
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing import image
import time
from gtts import gTTS
import tempfile
import base64

# Load the Keras model (.keras file)
model = tf.keras.models.load_model("Model.keras")

# Image parameters for prediction
img_height, img_width = 128, 128

def predict(img):
    # Start timer to track inference time
    start_time = time.time()
    
    # Preprocess the image
    img_array = image.img_to_array(img) / 255.0  # Normalize pixel values
    img_array = np.expand_dims(img_array, axis=0)  # Add batch dimension
    
    # Make prediction using the model
    prediction = model.predict(img_array)
    label = "The eyes are visibly closed, hinting at a danger ahead for the driver." if prediction[0] > 0.5 else "The eyes are visibly open, hinting at the driver's alertness."
    
    # Calculate inference time
    inference_time = time.time() - start_time
    
    return label, round(inference_time, 4)  # Return label and inference time

# Function to convert text to speech and auto-play
def speak_auto(text):
    tts = gTTS(text=text, lang='en')
    with tempfile.NamedTemporaryFile(delete=True, suffix=".mp3") as fp:
        tts.save(fp.name)
        audio_bytes = fp.read()
        b64 = base64.b64encode(audio_bytes).decode()
        audio_html = f"""
            <audio autoplay>
                <source src="data:audio/mp3;base64,{b64}" type="audio/mp3">
            </audio>
        """
        st.markdown(audio_html, unsafe_allow_html=True)

# Streamlit UI
st.title("Eye State Prediction")
st.write("Team 18: Sayandip Bhattacharyya, Purnendu Rudrapal, Sridatta Das, Sidhartha Karjee")

# Upload image
uploaded_image = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

if uploaded_image is not None:
    # Display uploaded image
    img = image.load_img(uploaded_image, target_size=(img_height, img_width))
    st.image(img, caption="Uploaded Image", use_container_width=True)

    # Make prediction
    label, inference_time = predict(img)

    # Display results
    st.write(f"Prediction: **{label}**")
    st.write(f"Inference Time: **{inference_time} seconds**")

    # Convert the result into speech and play it
    speak_auto(f"{label} Inference time: {inference_time} seconds.")