import streamlit as st
from transformers import pipeline
from PIL import Image
import requests
from io import BytesIO

# Set page configuration
st.set_page_config(
    page_title="Age Group Classification",
    page_icon="👶👧👩👵",
    layout="centered"
)

# Add title and description
st.title("Age Group Classification")
st.markdown("Upload an image of a person to classify their approximate age group using the `nateraw/vit-age-classifier` model.")

@st.cache_resource
def load_model():
    """Load the age classification model and cache it."""
    return pipeline("image-classification", model="nateraw/vit-age-classifier")

def load_image_from_url(url):
    """Load an image from a URL."""
    response = requests.get(url)
    img = Image.open(BytesIO(response.content))
    return img

# Load the model
pipe = load_model()

# Create file uploader
uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

# Function to get top prediction
def get_top_prediction(predictions):
    # Get the prediction with highest confidence
    top_prediction = max(predictions, key=lambda x: x['score'])
    return top_prediction

# Process the image and display results
if uploaded_file is not None:
    # Process uploaded image
    image = Image.open(uploaded_file)
    st.image(image, caption="Uploaded Image", use_container_width=True)
    
    with st.spinner("Analyzing age..."):
        predictions = pipe(image)
        top_pred = get_top_prediction(predictions)
        
    # Display result
    st.markdown("### Result:")
    st.metric(
        label="Predicted Age Range", 
        value=top_pred['label'], 
        delta=f"Confidence: {top_pred['score']:.2%}"
    )

elif 'example_loaded' in st.session_state and st.session_state.example_loaded:
    # Process example image
    with st.spinner("Loading example image..."):
        # Download and load the image properly
        image = load_image_from_url(st.session_state.example_image)
        st.image(image, caption="Example Image", use_container_width=True)
    
    with st.spinner("Analyzing age..."):
        # Pass the actual PIL Image object to the pipeline
        predictions = pipe(image)
        top_pred = get_top_prediction(predictions)
    
    # Display result
    st.markdown("### Result:")
    st.metric(
        label="Predicted Age Range", 
        value=top_pred['label'], 
        delta=f"Confidence: {top_pred['score']:.2%}"
    )

# Add information about the model
st.markdown("---")
st.markdown("### About the Model")
st.markdown("""
This app uses the `nateraw/vit-age-classifier` model from Hugging Face, which classifies 
images into age groups like "0-2", "3-9", "10-19", etc. The app displays only the most 
likely age range prediction with its confidence score.
""")