common_functions.py

import matplotlib.pyplot as plt
from matplotlib import cm
%matplotlib inline

# A function to plot a matrix
def plot_image(data, colorspace='binary'):
    fig = plt.figure(figsize=(10,10)) 
    ax = fig.add_subplot(111)
    ax.imshow(data, cmap=colorspace)
    
    
def normalize_array(array):
    array_min = array.min(axis=(0, 1), keepdims=True)
    array_max = array.max(axis=(0, 1), keepdims=True)
    normalized_array = (array - array_min)/(array_max - array_min)
    
    return normalized_array

def normalize_with_preset(array, max_value, min_value):
    normalized_array = (array - min_value)/(max_value - min_value)
    
    return normalized_array


def recreate_image(labels, w, h):
    """Recreate the (compressed) image from the code book & labels"""
    image = np.zeros((w, h))
    label_idx = 0
    for i in range(w):
        for j in range(h):
            image[i][j] = labels[label_idx]
            label_idx += 1
    return image

def overlay_images(image1, image2, colorspace1='binary', colorspace2='magma', transparency=0.5):
    plt.imshow(image1, cmap=colorspace1, interpolation='nearest')
    plt.imshow(image2, cmap=colorspace2, alpha=transparency, interpolation='bilinear')
    plt.show()
    
def flatten_normalize_neighbors(training_set, global_max=233, global_min=69):
    flattened_neighbors = []
    neighbors = training_set['neighbors']

    for i in range(len(neighbors)):
        numpy_array = np.array(neighbors[i]).reshape(-1, 1)
        normalized_neighbors = normalize_with_preset(numpy_array, global_max, global_min)
        flattened_neighbors.append(normalized_neighbors)
    
    return np.array(flattened_neighbors).reshape((len(flattened_neighbors), -1))

def extract_values_from_training_data_locations(training_data, image_channel):
    data_values = []
    
    for row in training_data:
        data_values.append(np.array([(image_channel[row[0], row[1]]), row[2]]))
        
    data_values = np.array(data_values)
    
    return data_values

def extract_neighborhood_values_from_training_data_locations(training_data, image_channel):
    data_values = []
    
    for row in training_data:
        data_values.append(np.array([(image_channel[row[0], row[1]]), row[2]]))
        
    data_values = np.array(data_values)
    
    return data_values

def flatten_neighbors(neighbors):
    data_values = []
    
    for neighbor in neighbors:
        #data_values.append(neighbor[0].reshape(-1, 1))
        data_values.append(neighbor[0].reshape(-1, 1))
    
    data_values = np.array(data_values)
    
    return data_values

def extract_labels_from_set(dataset):
    labels = []
    for neighbor in training_set:
        labels.append(neighbor[1])

    return np.array(labels).flatten()