push GLCM optimisé

TP.py

@@ -5,6 +5,7 @@ import pandas as pd
 import numpy as np
 from sklearn.metrics import accuracy_score
 from sklearn.model_selection import train_test_split
+from skimage.feature import graycomatrix, graycoprops
 from sklearn.naive_bayes import GaussianNB
 
 
@@ -30,10 +31,18 @@ def buildSampleFromPath(path1, path2, size=0):
     return S
 
 
-def computePixelBW_histo(image):
-    return image.convert("L").histogram()
+def computePixelBW_histo(image_gl):
+    return image_gl.histogram()
 
-def compute_glcm(image):
+def compute_glcm_caracteristics(image_gl):
+    image_arr= np.array(image_gl)
+    #print(image_arr.shape)
+    glcm = graycomatrix(image_arr, distances=[5], angles=[0], levels=256,
+                        symmetric=True, normed=True)
+    return [graycoprops(glcm, 'dissimilarity')[0, 0], graycoprops(glcm, 'correlation')[0, 0], graycoprops(glcm, 'contrast')[0, 0],
+            graycoprops(glcm, 'energy')[0, 0], graycoprops(glcm, 'homogeneity')[0, 0]]
+
+"""def compute_glcm(image):
     img_array = np.array(image)
     distance, angle = 5, 0
     glcm = np.zeros((256, 256))
@@ -48,12 +57,19 @@ def compute_glcm(image):
     return glcm
 
 def extract_data_glcm(glcm):
+    contrast, energy, homogeneity = 0, 0, 0
+    for i in range(glcm.shape[0]):
+        for j in range(glcm.shape[1]):
+            contrast += (i - j) ** 2 * glcm[i, j]
+            energy += glcm[i, j]**2
+            homogeneity += (glcm[i, j] / (1 + abs(i - j)))
+
     contrast = np.sum([[(i - j) ** 2 * glcm[i, j] for j in range(glcm.shape[1])] for i in range(glcm.shape[0])])
     energy = np.sum(glcm ** 2)
     homogeneity = np.sum(
         [[(glcm[i, j] / (1 + abs(i - j))) for j in range(glcm.shape[1])] for i in range(glcm.shape[0])])
 
-    return [contrast, energy, homogeneity]
+    return [contrast, energy, homogeneity]"""
 
 def computeDict(image_path, path, y_true_value, max_size: tuple):
     """
@@ -70,14 +86,15 @@ def computeDict(image_path, path, y_true_value, max_size: tuple):
     image = image.convert("RGB")
 
     resized = resizeImage(image, *max_size) # On ne stocke pas resized image, on calcule tout avant de l'oublier
-
+    image_gl = resized.convert("L")
 
     #print(computePixelBW_histo(resized))
     return {"name_path": full_path,
             #"resized_image": resized,
             "X_histo": computeHisto(resized),
             "X_pixelbw": computePixelBW_histo(resized),
-            "X_glcm_data": extract_data_glcm(compute_glcm(resized)),
+            #"X_glcm_data": extract_data_glcm(compute_glcm(resized)),
+            "X_glcm_data": compute_glcm_caracteristics(image_gl),
             "y_true_class": y_true_value,
             "y_predicted_class": None}