BETTER-NMA 1.0.5__py3-none-any.whl → 1.0.6__py3-none-any.whl

BETTER_NMA/utilss/classes/adversarial_dataset.py

@@ -1,6 +1,7 @@
 from sklearn.model_selection import train_test_split
 import numpy as np
 from .score_calculator import ScoreCalculator
+from ..photos_uitls import preprocess_numpy_image
 
 class AdversarialDataset:
     def __init__(self, model, clear_images, adversarial_images, Z_full, labels):
@@ -16,8 +17,9 @@ class AdversarialDataset:
         try:
             for image in self.clear_images[:50]:
                 # Add batch dimension for model prediction
-                image_batch = np.expand_dims(image, axis=0)
-                score = self.score_calculator.calculate_adversarial_score(self.model.predict(image_batch, verbose=0))
+                # image_batch = np.expand_dims(image, axis=0)
+                preprocessed_img = preprocess_numpy_image(self.model, image)
+                score = self.score_calculator.calculate_adversarial_score(self.model.predict(preprocessed_img, verbose=0))
                 scores.append(score)
                 labels.append(0)
         except Exception as e:
@@ -27,8 +29,9 @@ class AdversarialDataset:
         try:
             for adv_image in self.adversarial_images[:50]:
                 # Add batch dimension for model prediction
-                adv_image_batch = np.expand_dims(adv_image, axis=0)
-                score = self.score_calculator.calculate_adversarial_score(self.model.predict(adv_image_batch, verbose=0))
+                # adv_image_batch = np.expand_dims(adv_image, axis=0)
+                preprocessed_adv_img = preprocess_numpy_image(self.model, adv_image)
+                score = self.score_calculator.calculate_adversarial_score(self.model.predict(preprocessed_adv_img, verbose=0))
                 scores.append(score)
                 labels.append(1)
         except Exception as e:

BETTER_NMA/utilss/photos_uitls.py

@@ -69,4 +69,22 @@ def preprocess_image(model, image):
     preprocess_input = get_cached_preprocess_function(model)
     image_array = preprocess_input(np.array(image))
     image_preprocessed = np.expand_dims(image_array, axis=0)
+    return image_preprocessed
+
+def preprocess_numpy_image(model, image):
+    """
+    Preprocess a NumPy array image for the given model.
+    """
+    # if isinstance(image, list):
+    #     image = np.array(image)
+    if image.ndim == 3:
+        # If the image is 3D, add a batch dimension
+        image = np.expand_dims(image, axis=0)
+
+    # Get the appropriate preprocessing function for the model
+    preprocess_input = get_cached_preprocess_function(model)
+
+    # Apply the preprocessing function
+    image_preprocessed = preprocess_input(image)
+
     return image_preprocessed

{BETTER_NMA-1.0.5.dist-info → BETTER_NMA-1.0.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: BETTER_NMA
-Version: 1.0.5
+Version: 1.0.6
 Summary: NMA: Dendrogram-based model analysis, white-box testing, and adversarial detection
 Author: BETTER_XAI
 Author-email: BETTERXAI2025@gmail.com

{BETTER_NMA-1.0.5.dist-info → BETTER_NMA-1.0.6.dist-info}/RECORD

@@ -12,12 +12,12 @@ BETTER_NMA/train_adversarial_detector.py,sha256=nMaQ-Pm2vP84qNR1GoKQiVPpmMC3rdor
 BETTER_NMA/white_box_testing.py,sha256=VZ5pImXUOpM6jWMOoIkTWymwPCsev75zQ2SudSJ0frw,3539
 BETTER_NMA/utilss/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 BETTER_NMA/utilss/models_utils.py,sha256=gBXY2LFH4iR-2GZmHeUnnB5n9t3VdjIc9sugHDrD3AM,671
-BETTER_NMA/utilss/photos_uitls.py,sha256=wxmIIKFgAKYkcYaK95UMjtY-LZS6NDVveKrHBQV8Q70,3166
+BETTER_NMA/utilss/photos_uitls.py,sha256=liGx2exIfGMrJgUngFfCvDLWvoUnISqnwyaDNbUyA1Q,3734
 BETTER_NMA/utilss/photos_utils.py,sha256=4EjDHbMjrJ8P9y-X4H05P4wez4uKNit60UGnu3sKsys,4412
 BETTER_NMA/utilss/verbal_explanation.py,sha256=_hrYZUjBUYOfuGr7t5r-DACooR5d60dRtGfUj7FbeZw,549
 BETTER_NMA/utilss/wordnet_utils.py,sha256=77qcmEQH3Krd1T8dQY-IXVpaEgfwlw406XRk4zYsghw,9482
 BETTER_NMA/utilss/classes/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-BETTER_NMA/utilss/classes/adversarial_dataset.py,sha256=LKmyseQetyBdoKrl7Q4MaTc7EzMZcMgqNdtCsdAvKHA,2299
+BETTER_NMA/utilss/classes/adversarial_dataset.py,sha256=RGRaaDV9idjMp7GOsUUTVems_1GzswhaiK1Se7y5gFw,2528
 BETTER_NMA/utilss/classes/adversarial_detector.py,sha256=BE_SxNEwcvuHERBiefefOmk1k6NJSo6juehkAjkEHuQ,2331
 BETTER_NMA/utilss/classes/dendrogram.py,sha256=vtKBFfwzcz8k01Goc83pZlWC2pO86endTJURlkUWVQI,5141
 BETTER_NMA/utilss/classes/edges_dataframe.py,sha256=q-RQ6beOeZeIgdEzwi8T5Ag2NBFySv7-ITD5m989nl4,1896
@@ -33,7 +33,7 @@ BETTER_NMA/utilss/classes/preprocessing/z_builder.py,sha256=T8ETfL7mMOgEj7oYNsw6
 BETTER_NMA/utilss/enums/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 BETTER_NMA/utilss/enums/explanation_method.py,sha256=Ang-rjvxO4AJ1IH4mwS8sNpSwt9jn3PlqFbPPT-R9I8,150
 BETTER_NMA/utilss/enums/heap_types.py,sha256=0z1d2qu1ZCbpWRXKD1dTopn3M4G1CxRQW9HWxVxyPIA,88
-BETTER_NMA-1.0.5.dist-info/METADATA,sha256=oqz6IAT6kY-q94bKDTGFxSxAXI47J1P_GWcZQXKeaxM,5100
-BETTER_NMA-1.0.5.dist-info/WHEEL,sha256=PZUExdf71Ui_so67QXpySuHtCi3-J3wvF4ORK6k_S8U,91
-BETTER_NMA-1.0.5.dist-info/top_level.txt,sha256=SVRNqWPvCnynWVyXNAYnf9CSQIvMAvE6iyyiGHodQgY,11
-BETTER_NMA-1.0.5.dist-info/RECORD,,
+BETTER_NMA-1.0.6.dist-info/METADATA,sha256=M2D3uM-BAPpo-O7z7ft7HABhe7Oo3kxI8pdn3-lPkgA,5100
+BETTER_NMA-1.0.6.dist-info/WHEEL,sha256=PZUExdf71Ui_so67QXpySuHtCi3-J3wvF4ORK6k_S8U,91
+BETTER_NMA-1.0.6.dist-info/top_level.txt,sha256=SVRNqWPvCnynWVyXNAYnf9CSQIvMAvE6iyyiGHodQgY,11
+BETTER_NMA-1.0.6.dist-info/RECORD,,