app.py

import streamlit as st
import numpy as np
import pickle
from PIL import Image
from sklearn.neighbors import NearestNeighbors
import os
import tensorflow as tf
from tensorflow.keras.applications.resnet50 import ResNet50, preprocess_input
from tensorflow.keras.preprocessing import image
from tensorflow.keras.layers import GlobalMaxPool2D
from numpy.linalg import norm

# Başlık
st.title("👕 Fashion Product Recommender")
st.write("Lütfen bir ürün görseli yükleyin, size benzer ürünleri önerelim.")

# Modeli önbelleğe al
@st.cache_resource
def load_model():
    base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
    base_model.trainable = False
    model = tf.keras.Sequential([
        base_model,
        GlobalMaxPool2D()
    ])
    return model

# Özellik çıkaran fonksiyon
def extract_features(img_path, model):
    img = image.load_img(img_path, target_size=(224, 224))
    img_array = image.img_to_array(img)
    expanded = np.expand_dims(img_array, axis=0)
    preprocessed = preprocess_input(expanded)
    result = model.predict(preprocessed).flatten()
    normalized = result / norm(result)
    return normalized

# Görsel yükleyici
uploaded_file = st.file_uploader("Bir görsel seç (.jpg, .png)", type=["jpg", "png"])

if uploaded_file is not None:
    st.image(uploaded_file, caption="Yüklenen Ürün", use_column_width=True)
    
    # Geçici kaydet
    temp_path = "temp.jpg"
    with open(temp_path, "wb") as f:
        f.write(uploaded_file.read())

    # Model ve veriyi yükle
    model = load_model()
    try:
        feature_list = pickle.load(open("Images_features.pkl", "rb"))
        filenames = pickle.load(open("filenames.pkl", "rb"))
    except Exception as e:
        st.error("❌ Özellik dosyaları yüklenemedi. 'Images_features.pkl' ve 'filenames.pkl' dosyalarını kontrol edin.")
        st.stop()

    # Özellik çıkar
    try:
        input_feature = extract_features(temp_path, model)
        neighbors = NearestNeighbors(n_neighbors=6, algorithm='brute', metric='euclidean')
        neighbors.fit(feature_list)
        distances, indices = neighbors.kneighbors([input_feature])

        st.subheader("🔍 En Benzer Ürünler:")
        for idx in indices[0]:
            # Hugging Face zip destekli gösterim
            try:
                filename = filenames[idx].replace("images/", "")  # örn: '10640.jpg'
                st.image("images.zip@" + filename, width=200)
            except Exception as e:
                st.warning(f"Görsel gösterilemedi: {filenames[idx]}")
    except Exception as e:
        st.error(f"Bir hata oluştu: {str(e)}")