testing RBF and Linear model

source/main.py

@@ -1,10 +1,11 @@
-from utilities import *
-from preprocessing import *
-import segmentation
+from utilities import Utilities
+from preprocessing import Preprocessing
+from segmentation import *
 import svm
 from postprocessing import Postprocessing 
 import time
 import pandas as pd
+import matplotlib.pyplot as plt
 
 
 def compare_segmentation_methods():
@@ -55,34 +56,44 @@ def compare_segmentation_methods():
     # ISIC_0027861
 
 if __name__ == "__main__":
-
+        
     #col_names = ['f_a_0', 'f_a_1', 'f_a_2', 'f_a_3', 'f_b_0', 'f_c_0', 'f_c_1', 'f_c_2', 'f_c_3', 'f_c_4']
-
-    training_data = svm.Prediction.data_frames(pd.read_csv('data_set_v1.csv' , index_col=False))
-
-    print("training_data=", type(training_data))
 
-    model_prediction = svm.Prediction.grid_search(training_data)
+    training_data = svm.Prediction.data_frames(pd.read_csv('data_set_v1.csv' , index_col=0))
 
-    '''
-    data_set_path = "D:/Uni/WS 22-23/Digitale Bildverarbeitung/common_dataset/Dataset/" # Ghassan
-    #data_set_path = "C:/Users/ancik/Documents/GitHub/archive/HAM10000_images/"
+    # training_data = svm.Prediction.data_frames(pd.read_csv('data_set_v2.csv' , index_col=0))
 
+    # print("training_data=", type(training_data))
+
+
+    model_prediction = svm.Prediction.grid_search_RBF(training_data)    
+
+"""
+# Define Data set Path
+
+    # data_set_path = "D:/Uni/WS 22-23/Digitale Bildverarbeitung/common_dataset/Dataset/" # Ghassan
+    data_set_path = "C:/Users/Yazan/Desktop/DBV literature/Data/Data_set" # Yazan
+    #data_set_path = "C:/Users/ancik/Documents/GitHub/archive/HAM10000_images/" # Anna
+
+    # features, independent_features = svm.Prediction.feature_extraction(data_set_path)
     data_set = []
 
     # start timer
     start_time = time.process_time()
-
+    
     # Metadata loading
     # load HAM 10 000 dataset labels
-    dataset_metadata_path = "D:/Uni/WS 22-23/Digitale Bildverarbeitung/dataset/HAM10000_metadata.csv"
+
+    dataset_metadata_path = "C:/Users/Yazan/Desktop/DBV literature/Data/HAM10000_metadata.csv"
+
     # which labels from metadata we consider malign=positive=1 (others benign=0=negative)
-    list_of_malign_labels = ['mel', 'bcc'] # bcc rarely metastizes
+    list_of_malign_labels = ['mel'] # bcc rarely metastizes
     meta_data = Utilities.extract_labels_HAM10000(dataset_metadata_path, list_of_malign_labels)
-
+    
     # gen_file_names rename to generate_file_paths
+
     images_paths = Utilities.gen_file_names(data_set_path)
-
+    
     img_count = 0
     img_failed = 0
     for img_path in images_paths:
@@ -100,12 +111,14 @@ def compare_segmentation_methods():
         gamma_image = Preprocessing.gamma_correction(img, gamma=0.85)
         blured_img = Preprocessing.blur(gamma_image)  
 
-        binary_image = segmentation.NormalizedOtsuWithAdaptiveThresholding.segment(blured_img)
-
+        binary_image = NormalizedOtsuWithAdaptiveThresholding.segment(blured_img)
+        
         # feature extraction 
         longest_contour = Postprocessing.find_contours(binary_image)     
-        features = Postprocessing.feature_extractrion(img_number, longest_cntr=longest_contour, image_shape=binary_image.shape)
-
+        features , independent_features = Postprocessing.feature_extractrion(img_number, longest_cntr=longest_contour, image_shape=binary_image.shape)
+
+        print("Independent Features ", independent_features)
+
         img_feature_list = { 'img_number': img_number,
 
                             'metadata_label': meta_data[img_number], # 1 for malign etc. positive, 0 for benign etc. negative
@@ -121,26 +134,31 @@ def compare_segmentation_methods():
                             'f_c_1':features[2][1],
                             'f_c_2':features[2][2],
                             'f_c_3':features[2][3],
-                            'f_c_4':features[2][4]                            
+                            'f_c_4':features[2][4],
+
+                            'ind_0':independent_features[0],
+                            'ind_1':independent_features[1],
+                            'ind_2':independent_features[2],
+                            'ind_3':independent_features[3],
+                            'ind_4':independent_features[4]
+
                             }
         if (None in img_feature_list.values()): img_failed += 1
 
         data_set.append(img_feature_list)
 
-    Utilities.save_dataset(dataset=data_set, file_path="./data_set.csv", only_succesfull=True)
+        Utilities.save_dataset(dataset=data_set, file_path="./data_set_v2.csv", only_succesfull=True)
 
-    end_time = time.process_time()
+        end_time = time.process_time()
 
-    total_time = (end_time - start_time)*1000 # in millis
+        total_time = (end_time - start_time)*1000 # in millis
 
-    avg_time = total_time / img_count
-
-    print("total_time = %.0f min" % (total_time/1000/60))
-    print("avg_time = %.0f ms per image" % avg_time)
-    print("img_failed: %d ... %.1f%% of total images" %(img_failed, img_failed/img_count*100))
-    print("img_count", img_count)
-    '''
-    # print("complete_data_set=", data_set)
+        avg_time = total_time / img_count
+
+        print("total_time = %.0f min" % (total_time/1000/60))
+        print("avg_time = %.0f ms per image" % avg_time)
+        print("img_failed: %d ... %.1f%% of total images" %(img_failed, img_failed/img_count*100))
+        print("img_count", img_count)
+"""
 
-
-
+

source/postprocessing.py

@@ -36,9 +36,9 @@ def feature_extractrion(image_number, longest_cntr, image_shape, open_contour_th
         #print("longest_cntr=", longest_cntr)
 
         first_tuple = (None, None, None, None)
-        ghassan_tuple = (None)
+        ellipse_tuple = (None)
         third_tuple = (None, None, None, None, None)
-
+        independent_features = (None, None, None, None, None)
         if longest_cntr is not None:
 
             largest_area = cv2.contourArea(longest_cntr)
@@ -75,6 +75,8 @@ def feature_extractrion(image_number, longest_cntr, image_shape, open_contour_th
 
                 '''since just closed contours are considered, the param @closed always set to True '''                
                 perimeter = cv2.arcLength(longest_cntr, True)
+
+
 
                 ''' Paper: Melanoma Skin Cancer Detection Using Image Processing and Machine Learning Techniques '''                                    
                 ir_A = perimeter / largest_area                                     
@@ -93,9 +95,10 @@ def feature_extractrion(image_number, longest_cntr, image_shape, open_contour_th
 
                 cv2.ellipse(external_contours,ellipse,(255),2)
 
-                ghassan_tuple = (ellipse_irrigularity)
+                ellipse_tuple = (ellipse_irrigularity)
 
-                ''' Paper: Computer aided Melanoma skin cancer detection using Image Processing ..'''            
+                ''' Paper: Computer aided Melanoma skin cancer detection using Image Processing ..'''     
+
                 Circularity_indx = (4*largest_area*np.pi) / perimeter**2
 
                 ir_A = perimeter / largest_area
@@ -110,6 +113,8 @@ def feature_extractrion(image_number, longest_cntr, image_shape, open_contour_th
 
                 # cv2.ellipse(external_contours,ellipse,255,2)
                 cv2.circle(external_contours, (com_x, com_y), 5, 255, -1)
+
+                independent_features = (perimeter,largest_area,minor_diameter,major_diameter,ellipse_irrigularity)
 
         else:
             #cv2.putText(external_contours,"Unable to detect closed contours!",(50,250),
@@ -118,4 +123,4 @@ def feature_extractrion(image_number, longest_cntr, image_shape, open_contour_th
             print("Unable to detect closed contours to this image:!" + str(image_number))   
 
 
-        return [first_tuple, ghassan_tuple, third_tuple]
+        return [first_tuple, ellipse_tuple, third_tuple] , independent_features