Neural-Network/run.py at main · YonatanVigiser/Neural-Network · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
import DrawingPad
from NeuralNetwork import NeuralNetwork
import numpy as np
import tensorflow as tf
from PIL import Image, ImageOps, ImageFilter
# import matplotlib.pyplot as plt

network = NeuralNetwork([784, 64, 32, 10])
(train_images, train_labels), (test_images, test_labels) = tf.keras.datasets.mnist.load_data()


def load_test_image(image_number):
    arr = np.array(test_images[image_number])
    arr = arr.tolist()
    res = []

    for n in arr:
        for x in n:
            res.append(x / 255)

    return res


def load_train_image(image_number):
    arr = np.array(train_images[image_number])
    arr = arr.tolist()
    res = []

    for n in arr:
        for x in n:
            res.append(x / 255)

    return res


def paint():
    DrawingPad.DrawingPad.run()
    digit = input("\nPlease enter the digit you have written: ")
    while not digit.isdigit():
        print("Please enter a digit! Try again.")
        digit = input("Please enter the digit you have written: ")

    drawn_image = Image.open('digit.png')
    drawn_image = drawn_image.resize((28, 28))
    drawn_image = drawn_image.convert('L')
    drawn_image = ImageOps.invert(drawn_image)
    drawn_image = drawn_image.filter(ImageFilter.GaussianBlur(radius=1))
    array = np.array(drawn_image) / 255.0
    array = array.reshape((1, 28, 28, 1))
    array = array.flatten().tolist()

    # To display the imagee drawn or any other image:
    # plt.imshow(image, cmap='gray')
    # plt.title(f"Label: {digit}")
    # plt.axis('off')
    # plt.show()

    result = network.run_network(array)

    if result[-1].index(max(result[-1])) == int(digit):
        print(f"\nGuessed correctly! ({round(max(result[-1]) * 100, 2)}% confidence)")
    else:
        print(f"\nGuessed incorrectly ({result[-1].index(max(result[-1]))})! "
              f"({round(max(result[-1]) * 100, 2)}% confidence)")
    again = input("\nDo you want to draw again? (Y/n): ")
    if again.upper() == "Y":
        paint()


training_count = 60000
testing_count = 10000
learning_rate = 0.1


train_images_result = []

for i in range(training_count):
    image = load_train_image(i)
    train_images_result.append(image)

test_images_result = []

for i in range(testing_count):
    image = load_test_image(i)
    test_images_result.append(image)

train = input("Do you want to use an exsisting saved model? (Y/n): ")
print()

if train.upper() == "Y":
    weights_file = open("weights.txt", "r")
    network.weights = eval(weights_file.read())
    weights_file.close()
    biases_file = open("biases.txt", "r")
    network.biases = eval(biases_file.read())
    biases_file.close()


else:
    network.train(train_images_result, train_labels[:training_count], learning_rate)

do_paint = input("Do you want to to paint the number? (Y/n): ")
if do_paint.upper() == "Y":
    paint()
else:
    network.test(test_images_result, test_labels[:testing_count])


if train.upper() != "Y":
    save = input("\nDo you want to save the network values? (Y/n): ")
    if save.upper() == "Y":
        weights_file = open("weights.txt", "w")
        weights_file.write(str(network.weights))
        weights_file.close()
        biases_file = open("biases.txt", "w")
        biases_file.write(str(network.biases))
        biases_file.close()