Merge pull request #3 from DerikVo/eda_create_module_avg_pixels

Eda create module avg pixels

Author: DerikVo

Created_images/Pituitary contrast.png

@@ -2,32 +2,58 @@
 from math import ceil
 import os
 import random
+import numpy as np
 from tensorflow.keras.layers.experimental import preprocessing
 from tensorflow.keras.preprocessing.image import img_to_array, load_img
 from tensorflow.keras.preprocessing import image_dataset_from_directory
 from keras.preprocessing.image import ImageDataGenerator
 from sklearn.decomposition import PCA
 import matplotlib.pyplot as plt
-import numpy as np
-np.random.seed(42)
-from tensorflow.keras.utils import set_random_seed
-set_random_seed(42)
 
-def folders(dataset):
+'''
+The concept of this code was learned from General Assembly's Data science Immersive's Excel lab excercise
+The concepts has been adapted to identify file paths ways to work with the data
+''' 
+def folders(dataset = 'Training'):
+    '''
+    Finds the folders within either the 'Training' dataset or the 'Testing' data set.
+    =============================================================================    
+    Keyword arguments:
+    dataset -- The main folder either Train and Test folder (default = Training)
+    
+    =================================================
+    Example:
+    train_folders = eda.folders(train_dataset)
+    test_folders = eda.folders(test_dataset)
+    '''
+    #sets the path to the Testing and training folders
     path = f'../Images/{dataset}'
-    images_dir = [x for x in os.listdir(path)]
+    #lists the classification folders within the dataset
     folders= os.listdir(path)
     return folders
 
-def image_len(dataset, folders):
+
+'''
+
+The concept of this code was learned from General Assembly's Data science Immersive's Excel lab excercise during the 2023-03 to 2023-06 cohort.
+The concepts has been adapted to identify file paths ways to work with the data
+'''
+def image_len(folders, dataset ='Training'):
     '''
-    This code takes in the list directory of the folder containing the classification folders. And the dataset.
-    this code was heavily inspired by this project: https://github.com/DerikVo/DSI_project_4_plant_disease/blob/main/notebooks/01_Potato_PlantVillageEDA.ipynb
-    Has since been adapted to work with a jupyter notebook
-    TODO:convert all image eda into a class/method script
+    
+    List subfolders with the main folder containing the classification folders for each image set. As well as shows a random image from the classification
+    =============================================================================    
+    Keyword arguments:
+    folders -- The sub folder containing the classifcation for tumor type ( 'glioma', 'meningioma', 'notumor', 'pituitary' )
+    dataset -- The main folder either Training and Testing folder (default = Training)
+    
+    ===============================================
+    Example:
+    eda.image_len(train_folders)
     '''
-    #loop to each sub folder so we can get the class sizes
+    #sets the path to the Testing and training folders
     path = f'../Images/{dataset}'
+    #loop through each classification folder
     for i in folders:
         #get number of images of folder
         num = len(os.listdir(f'{path}/{i}'))
@@ -39,51 +65,132 @@ def image_len(dataset, folders):
         image_name = (os.listdir(f'{path}/{i}')[rand])
         #assigns the file path to the image
         image = load_img(f'{path}/{i}/{image_name}')
-        #shows the image
+        #shows the image for the classification for reference
         plt.title(f'{image_name}')
         plt.imshow(image)
         plt.axis('off')
         plt.show()
-class Image:
-    def __init__(self, dataset, sub_folder):
-        #learned i didnt need a comma because that creates a tuple: https://stackoverflow.com/questions/39192261/class-init-takes-parameters-but-turns-them-into-tuples-for-some-reason
-        self.dataset = dataset
-        self.sub_folder = sub_folder
-    
-    def avg_images(self):
-      '''
-      This function takes two arguments the dataset: training or testing, and the sub_folder for the type of tumor e.g. ['glioma', 'meningioma', 'notumor', 'pituitary']
-      This function is used to find the average pixel values of each class
-      The purpose is to find if there is a difference in each class
-      '''
-      #assign the path in the function for readability and understanding
-      #assign the sub folder (class name) that was passed to the function
-      path = (f'../Images/{self.dataset}')
-      class_name = self.sub_folder
-      batch_size = 32  # Modify this to suit your needs
-      #instantiate ImageDataGenerator
-      datagen = ImageDataGenerator(rescale=1./255)  # normalize pixel values to [0,1]
-      #get the images from the directory
-      generator = datagen.flow_from_directory(path,
-                                              classes=[class_name],
-                                              class_mode=None,
-                                              color_mode='grayscale',
-                                              target_size=(256, 256),
-                                              batch_size=batch_size)
-      n_samples = generator.samples
-      average_image = np.zeros((256, 256, 1))
+'''
+This portion uses code from a previous project from this [notebook](https://github.com/DerikVo/DSI_project_4_plant_disease/blob/main/notebooks/01_Potato_PlantVillageEDA.ipynb).
+The code was originally developed by chat GPT 4 with the prompt:  "I have an image data set that I want to do EDA on. How can I average out the pixel values of all the images in a class. python keras."
 
-      for i in range(n_samples // batch_size):  # Integer division to avoid partial batches
-          images = next(generator)
-          average_image += np.sum(images, axis=0)
+This function takes two arguments the dataset: training or testing, and the sub_folder for the type of tumor e.g. ['glioma', 'meningioma', 'notumor', 'pituitary']
+This function is used to find the average pixel values of each class
+The purpose is to find if there is a difference in each class
 
-      average_image /= n_samples
-      return average_image
+'''
+def avg_images(folders, dataset = 'Training'):
+    '''
+
+    This function is used to find the average pixel value of each class
     
+    Users will need to assign the images to a variable. 
+    For example:
+    meningioma_tumor = eda.avg_images('meningioma')
+    =============================================================================    
+    Keyword arguments:
+    folders -- The sub folder containing the classifcation for tumor type ( 'glioma', 'meningioma', 'notumor', 'pituitary' )
+    dataset -- The main folder either Train and Test folder (default = Training)
+    '''
+    #sets the path to the Testing and training folders
+    path = (f'../Images/{dataset}')
+  
+    class_name = folders
+    batch_size = 32  # Modify this to suit your needs
+    #instantiate ImageDataGenerator
+    datagen = ImageDataGenerator(rescale=1./255)  # normalize pixel values to [0,1]
+    #get the images from the directory
+    generator = datagen.flow_from_directory(path,
+                                          classes=[class_name],
+                                          class_mode=None,
+                                          color_mode='grayscale',
+                                          target_size=(256, 256),
+                                          batch_size=batch_size)
+    n_samples = generator.samples
+    average_image = np.zeros((256, 256, 1))
+
+    for i in range(n_samples // batch_size):  # Integer division to avoid partial batches
+        images = next(generator)
+        average_image += np.sum(images, axis=0)
+
+    average_image /= n_samples
+    return average_image
+
+'''
+This portion uses code from a previous project from this [notebook](https://github.com/DerikVo/DSI_project_4_plant_disease/blob/main/notebooks/01_Potato_PlantVillageEDA.ipynb). The concept was originally developed by [Yasser Siddiqui]([email protected]) and has been adapted to use with this notebook.
+
+This function is used to find the differences of the average pixel value between each class 'glioma', 'meningioma', and 'pituitary' compared to 'notumor'. 
+These different characteristics can help us understand how the classes are unique when compared to not having a tumor.
+If there are significant differences we can better interpret our model.
+'''
 def image_contrast(comparision, base_image):
+    '''
+    This function finds the differences between the pixel averages of two classes to identify how the model can differentiate classes
+    Users will need to have ran the avg_images function for each class before running the image_contrats function.
+    
+    Users will need to assign the images to a variable. 
+    For example:
+    meningioma_contrast = eda.image_contrast(meningioma_tumor, notumor)
+    =============================================================================    
+    Keyword arguments:
+    comparision -- The image that contains the type of tumor ( 'glioma', 'meningioma', 'pituitary' )
+    base_image -- The image youre comparing against ('notumor')
+    '''
     # we need to rescale the contrasts
     image = base_image - comparision
-    image -= image.min() # subtract minimum 
-    image /= image.max() # divide by new max
+    # subtract minimum 
+    image -= image.min()
+    # divide by new max
+    image /= image.max()
+
+    return image
+'''
+refernced resources for colorblindness accomadation:
+https://matplotlib.org/3.1.0/tutorials/colors/colormaps.html#colorblindness
 
-    return image
+'''
+def display_image(image, title):
+    '''
+
+    This function is to plot the three of the same images with different supports for varying color blindness 
+    This function can be used for both the image_contrast and avg_images funcitons.
+    
+    Users will need to assign the images to a variable. 
+    For example:
+    meningioma_contrast = eda.image_contrast(meningioma_tumor, notumor)
+    =============================================================================    
+    Keyword arguments:
+    Image -- The image you want to display
+    Title -- The title of the image. This will be used for saving the figure as well
+    '''
+    
+    #sets up the figure for the subplots
+    fig, ax = plt.subplots(1,3, figsize = (10,10))
+    #plots the title for the color map
+    plt.suptitle(f'Pixel average: {title}', y = .75, fontsize = 22)
+    #plotting the images
+    ax[0].imshow(image)
+    #uses the default Viridis color map; default colorblind friendly
+    ax[0].set_title('Viridis', fontsize=18)
+    #turns of axis bcause we only want the image
+    ax[0].axis('off')
+   
+    #plotting the images
+    ax[1].imshow(image, cmap='cividis')
+    #plots the title for the color map
+    ax[1].set_title('cividis', fontsize=18)
+    #turns of axis bcause we only want the image
+    ax[1].axis('off')
+    
+    #plotting the images
+    ax[2].imshow(image, cmap='magma')
+    #plots the title for the color map
+    ax[2].set_title('magma', fontsize=18)
+    #turns of axis bcause we only want the image
+    ax[2].axis('off')
+    
+    plt.tight_layout()
+    #saves the image
+    plt.savefig(f'../Created_images/{title}.png')
+    #displays the image in the notebook
+    plt.show()