import streamlit as st
import numpy as np
import pickle
import cv2

# -------------------------------
# Page Config
# -------------------------------
st.set_page_config(page_title="🚗 Car Brand Logo Classifier")

st.title("🚗 Car Brand Logo Classifier")

# -------------------------------
# Load the full pipeline (Scaler + PCA + Classifier)
# -------------------------------
with open("car_logo_pipeline.pkl", "rb") as f:
    pipeline = pickle.load(f)

# -------------------------------
# File Uploader
# -------------------------------
uploaded_file = st.file_uploader("Upload a car logo image", type=["jpg", "jpeg", "png"])

if uploaded_file:
    # Read uploaded file directly into OpenCV format (BGR, uint8) just like training
    file_bytes = np.asarray(bytearray(uploaded_file.read()), dtype=np.uint8)
    img_array = cv2.imdecode(file_bytes, cv2.IMREAD_COLOR)

    # Display image in Streamlit (convert to RGB for display only)
    st.image(cv2.cvtColor(img_array, cv2.COLOR_BGR2RGB), caption="Uploaded Logo", use_column_width=True)

    # Resize to match training preprocessing
    resized_img = cv2.resize(img_array, (300, 300))

    # Flatten for model input
    flat_img = resized_img.flatten().reshape(1, -1)

    # Predict button
    if st.button("🔍 Predict Brand"):
        prediction = pipeline.predict(flat_img)[0]
        st.markdown(f"## ✅ Predicted Brand: `{prediction}`")