Update app.py

app.py

@@ -1,93 +1,109 @@
-import streamlit as st
-import os
-import numpy as np
-import pickle as pkl
-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 sklearn.neighbors import NearestNeighbors
-from numpy.linalg import norm
-from PIL import Image
-
-# Sayfa Ayarları ve Ortalama
-st.set_page_config(page_title="Moda Öneri Sistemi", layout="centered")
-
-st.markdown("""
-    <style>
-    .stTitle, .stSubheader, p { text-align: center; }
-    .stImage { display: flex; justify-content: center; }
-    </style>
-    """, unsafe_allow_html=True)
-
-st.title('🛍️ Moda Öneri Sistemi')
-
-@st.cache_resource
-def load_data():
-    base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
-    base_model.trainable = False
-    model = tf.keras.models.Sequential([base_model, GlobalMaxPool2D()])
-    
-    # Pickle dosyalarını arama
-    pkl_options = ['Images_features.pkl', 'embeddings.pkl', 'src/Images_features.pkl', 'src/embeddings.pkl']
-    features_file = next((f for f in pkl_options if os.path.exists(f)), None)
-    
-    fn_options = ['filenames.pkl', 'src/filenames.pkl']
-    filenames_file = next((f for f in fn_options if os.path.exists(f)), None)
-
-    if not features_file or not filenames_file:
-        st.error("HATA: .pkl dosyaları bulunamadı!")
-        return None, None, None
-
-    features = np.array(pkl.load(open(features_file, 'rb')))
-    filenames = pkl.load(open(filenames_file, 'rb'))
-    return model, features, filenames
-
-model, feature_list, filenames = load_data()
-
-def extract_features(img_path, model):
-    img = image.load_img(img_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
-
-uploaded_file = st.file_uploader("Kıyafet resmi yükleyin...", type=['jpg', 'png', 'jpeg'])
-
-if uploaded_file is not None and model is not None:
-    col1, col2, col3 = st.columns([1, 2, 1])
-    with col2:
-        display_image = Image.open(uploaded_file)
-        st.image(display_image, use_container_width=True, caption='Yüklenen Resim')
-    
-    temp_path = "temp_upload.jpg"
-    with open(temp_path, "wb") as f:
-        f.write(uploaded_file.getbuffer())
-    
-    input_features = extract_features(temp_path, model)
-    neighbors = NearestNeighbors(n_neighbors=6, algorithm='brute', metric='euclidean')
-    neighbors.fit(feature_list)
-    distances, indices = neighbors.kneighbors([input_features])
-    
-    st.markdown("---")
-    st.subheader('✨ Benzer Ürünler')
-    
-    cols = st.columns(5)
-    for i in range(1, 6):
-        with cols[i-1]:
-            # DOSYA YOLU HATASINI ÇÖZEN KISIM:
-            raw_path = filenames[indices[0][i]].replace('\\', '/')
-            img_name = os.path.basename(raw_path) # Örn: '1528.jpg'
-            
-            # Hugging Face'te resimlerin olduğu muhtemel yerler
-            p1 = os.path.join('images', img_name)
-            p2 = os.path.join('src', 'images', img_name)
-            
-            if os.path.exists(p1):
-                st.image(p1, use_container_width=True)
-            elif os.path.exists(p2):
-                st.image(p2, use_container_width=True)
-            else:
-                st.error(f"Eksik: {img_name}")
+Haklısın, parça parça eklemeler yapınca kafa karışması çok normal. Şimdi her şeyi tek bir pakette birleştirelim.
+
+Hugging Face'te çalışan, dosya yollarını otomatik düzelten ve büyük/küçük harf hatalarını (Linux kaynaklı) engelleyen en sağlam kodu aşağıya yazıyorum.
+
+1. app.py (Tüm Kod)
+Bu kodu kopyala ve Hugging Face'teki app.py dosyasının içine tamamen yapıştır (eskisini sil):
+
+Python
+import streamlit as st
+import os
+import numpy as np
+import pickle as pkl
+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 sklearn.neighbors import NearestNeighbors
+from numpy.linalg import norm
+from PIL import Image
+
+# Sayfa tasarımı
+st.set_page_config(page_title="Moda Öneri Sistemi", layout="centered")
+
+st.markdown("""
+    <style>
+    .stTitle, .stSubheader, p { text-align: center; }
+    .stImage { display: flex; justify-content: center; }
+    </style>
+    """, unsafe_allow_html=True)
+
+st.title('🛍️ Moda Öneri Sistemi')
+
+# Model ve verileri yükle
+@st.cache_resource
+def load_data():
+    # ResNet50 modelini hazırla
+    base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
+    base_model.trainable = False
+    model = tf.keras.models.Sequential([base_model, GlobalMaxPool2D()])
+    
+    # Pickle dosyalarını kontrol et (Ana dizinde arar)
+    try:
+        features = np.array(pkl.load(open('Images_features.pkl', 'rb')))
+        filenames = pkl.load(open('filenames.pkl', 'rb'))
+        return model, features, filenames
+    except FileNotFoundError:
+        st.error("HATA: .pkl dosyaları ana dizinde bulunamadı!")
+        return None, None, None
+
+model, feature_list, filenames = load_data()
+
+def extract_features(img_path, model):
+    img = image.load_img(img_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
+
+# Dosya yükleme alanı
+uploaded_file = st.file_uploader("Kıyafet resmi seçin...", type=['jpg', 'png', 'jpeg'])
+
+if uploaded_file is not None and model is not None:
+    # Seçilen resmi göster
+    col1, col2, col3 = st.columns([1, 2, 1])
+    with col2:
+        display_image = Image.open(uploaded_file)
+        st.image(display_image, use_container_width=True, caption='Yüklediğiniz Resim')
+    
+    # Geçici olarak kaydet
+    temp_path = "temp_upload.jpg"
+    with open(temp_path, "wb") as f:
+        f.write(uploaded_file.getbuffer())
+    
+    # Benzerleri bul
+    with st.spinner('Öneriler hazırlanıyor...'):
+        input_features = extract_features(temp_path, model)
+        neighbors = NearestNeighbors(n_neighbors=6, algorithm='brute', metric='euclidean')
+        neighbors.fit(feature_list)
+        distances, indices = neighbors.kneighbors([input_features])
+    
+    st.markdown("---")
+    st.subheader('✨ Benzer Ürünler')
+    
+    cols = st.columns(5)
+    
+    # Resim klasörünü tara (Büyük/küçük harf duyarlılığı için)
+    image_folder = 'images'
+    if os.path.exists(image_folder):
+        available_files = os.listdir(image_folder)
+        # Dosya adlarını küçük harfe çevirerek bir sözlük oluştur (Hızlı arama için)
+        file_map = {f.lower(): f for f in available_files}
+    else:
+        file_map = {}
+        st.error("'images' klasörü bulunamadı!")
+
+    for i in range(1, 6):
+        with cols[i-1]:
+            # Pickle içindeki yolu temizle (Windows yollarını Linux'a çevir)
+            raw_path = filenames[indices[0][i]].replace('\\', '/')
+            target_name = os.path.basename(raw_path).lower()
+            
+            # Klasörde bu resim var mı bak
+            if target_name in file_map:
+                final_path = os.path.join(image_folder, file_map[target_name])
+                st.image(final_path, use_container_width=True)
+            else:
+                st.warning(f"Bulunamadı: {target_name}")