import streamlit as st
import pickle as pkl
from PIL import Image
from sklearn.neighbors import NearestNeighbors
from tensorflow.keras.models import load_model
from tensorflow.keras.applications.resnet50 import preprocess_input
from tensorflow.keras.preprocessing import image
import os
from numpy.linalg import norm
import numpy as np

# Load necessary files and model
filenames = pkl.load(open('filenames2.pkl', 'rb'))
Image_features = pkl.load(open('Images_features2.pkl', 'rb'))
neighbors = NearestNeighbors(n_neighbors=6, algorithm='brute', metric='euclidean')
neighbors.fit(Image_features)
model = load_model('model.h5')

# Feature extraction function
def extract_features_from_images(image_path, model):
    img = image.load_img(image_path, target_size=(224, 224))
    img_array = image.img_to_array(img)
    img_expand_dim = np.expand_dims(img_array, axis=0)
    img_preprocess = preprocess_input(img_expand_dim)
    result = model.predict(img_preprocess).flatten()
    norm_result = result / norm(result)
    return norm_result

# Custom CSS for background and title
st.markdown(
    """
    <style>
    body {
        background-color: #f0f8ff; /* Soft light blue background */
    }
    .title {
        color: #1f7a8c; /* Teal color */
        font-family: 'Arial Black', sans-serif;
        text-align: center;
    }
    </style>
    """,
    unsafe_allow_html=True,
)

# Streamlit UI
st.markdown('<h1 class="title">Fashion Deep Learning Classification 👗</h1>', unsafe_allow_html=True)
st.write('Upload a clothing image, and the model recommends similar products.')

file = st.file_uploader('Upload an Image', type=['jpg', 'jpeg', 'png'])

if file is not None:
    # Display uploaded image with reduced size
    img = Image.open(file)
    st.image(img, caption='Uploaded Image', width=200)

    # Extract features from the uploaded image
    input_image = extract_features_from_images(file, model)

    # Get recommendations
    distance, indices = neighbors.kneighbors([input_image])
    st.subheader("Recommended Products:")

    # Display recommendations in a grid layout
    cols = st.columns(3)  # 3 images per row

    for i, idx in enumerate(indices[0]):
        product_path = filenames[idx]
        st.write(product_path)
        product_image = Image.open(product_path)
        
        # Assign columns dynamically
        with cols[i % 3]:
            st.image(product_image, caption=f"Product {i + 1}", use_column_width=True)