Update pages/Entorno de Ejecución.py

pages/Entorno de Ejecución.py

@@ -10,7 +10,7 @@ import sklearn
 from PIL import Image 
 from sklearn.decomposition import PCA
 from tensorflow.keras.models import load_model
-from transformers import AutoConfig, AutoModel, pipeline, Dinov2Model#, AutoProcessor, AutoModelForZeroShotImageClassification
+from transformers import AutoConfig, AutoModel, pipeline, Dinov2Model
 
 
 token = os.environ['token']
@@ -20,15 +20,17 @@ st.set_page_config(
     layout = 'wide',
     menu_items = {
         "About" : 'Proyecto ideado para la investigación de "Clasificación de imágenes de una sola clase con algortimos de Inteligencia Artificial".',
-        "Report a Bug" : 'https://docs.google.com/forms/d/e/1FAIpQLScH0ZxAV8aSqs7TPYi86u0nkxvQG3iuHCStWNB-BoQnSW2V0g/viewform?usp=sf_link'
+        "Report a Bug" : 'contact@patacotron.tech'
     }
 )
 
 st.sidebar.write("contact@patacotron.tech")
-already_excuted = False
 cnn, vit, zero_shot, classic_ml = st.tabs(["CNN", "ViT", "Zero-Shot", "Machine Learning Clásico"])
 classic_ml_root = "/home/user/app/classicML"
-pca = joblib.load(os.path.join(classic_ml_root, "pca_model.pkl"))
+
+@st.cache_resource
+def load_pca():
+    pca = joblib.load(os.path.join(classic_ml_root, "pca_model.pkl"))
 
 def _predict(_model_list, _weights, _img): 
     y_gorrito = 0
@@ -38,14 +40,24 @@ def _predict(_model_list, _weights, _img):
         y_gorrito += tf.cast(model(tf.expand_dims(img/255., 0)), dtype=tf.float32)*weight
     return [y_gorrito / sum(_weights), raw_img]
 
-def classic_ml_prediction(clfs, _img):
+def _pca_predict(models, _img):
     y_gorrito = 0
     raw_img = cv2.cvtColor(_img, cv2.COLOR_BGR2RGB)
-    img = cv2.resize(_img, (IMAGE_WIDTH, IMAGE_HEIGHT)).flatten()
-    data = pca.fit_transform(img)
-    for clf in clfs:
-        y_gorrito += clf.predict(data)
-    return [y_gorrito / len(clfs), raw_img] 
+    img = cv2.resize(_img, (IMAGE_WIDTH, IMAGE_HEIGHT))
+    fl_img =[img.flatten()]
+    data = pca.transform(fl_img)
+    for model in models:
+        y_gorrito += model.predict_proba(data)[0][Categories.index("Patacon-True")]
+    return [y_gorrito / len(models), raw_img]
+
+#def classic_ml_prediction(clfs, _img):
+#    y_gorrito = 0
+#    raw_img = cv2.cvtColor(_img, cv2.COLOR_BGR2RGB)
+#    img = cv2.resize(_img, (IMAGE_WIDTH, IMAGE_HEIGHT)).flatten()
+#    data = pca.transform(img.reshape(1, -1))
+#    for clf in clfs:
+#        y_gorrito += clf.predict(data)
+#    return [y_gorrito / len(clfs), raw_img] 
 
 def preprocess(file_uploader, module = 'cv2'): #makes the uploaded image readable
     img = np.frombuffer(uploaded_file.read(), np.uint8)
@@ -281,6 +293,8 @@ with zero_shot:
                 st.write("Asegúrate de haber subido correctamente la imagen.")
         
 with classic_ml:
+    
+    load_pca()
     col_a, col_b = st.columns(2)
     with col_a:
     
@@ -311,7 +325,8 @@ with classic_ml:
                 with st.spinner('Cargando predicción...'):
 
                     img = preprocess(uploaded_file)
-                    y_gorrito, raw_img =  classic_ml_prediction(model_choice, img)
+                    model_list = [joblib.load(directory) for directory in model_choice]
+                    y_gorrito, raw_img =  _pca_predict(model_list, img)
                     
                     if round(float(y_gorrito*100)) >= threshold:
                         st.success("¡Patacón Detectado!")