FaceRecognition/visualizer.py at master · finked/FaceRecognition · 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
"""
visualizer

methods to get a better understanding of the data we are using
"""

# TODO:
# - showImage / showImages: generalize magic numbers

import numpy as np
import matplotlib.pyplot as plt

def showImage(image, prediction=None, dimensions=None, axis=None, cmap='gray'):
    """
    show one image

    image        - pixel data to draw the image from
    [prediction] - facial keypoints to draw over the image
    [dimensions] - dimensions of the image to reshape. If not set, a square
                   image is assumed
    [axis]       - an subplot object to draw into. If not set, the image is
                   plotted directly
    [cmap]       - color map. Default is gray.
    """

    if not dimensions:
        dim = int(np.sqrt(len(image)))
        dimensions = (dim, dim)

    img = image.reshape(dimensions[0], dimensions[1])

    if axis:
        axis.imshow(img, cmap=cmap)
        if prediction is not None:
            axis.scatter(prediction[0::2] * 48 + 48,
                         prediction[1::2] * 48 + 48,
                         marker='x',
                         s=80
                        )
    else:
        plt.imshow(img, cmap=cmap)
        if prediction is not None:
            plt.scatter(prediction[0::2] * 48 + 48,
                        prediction[1::2] * 48 + 48,
                        marker='x',
                        s=80
                       )
        plt.show()


def showImages(images, predictions=None, imagesInRow=4):
    """
    show multiple images

    this creates a grid with the images.
    The standard number of Images in a row is 4.
    If `imagesInRow` is 0, then all images are printed in one row.
    If predictions are given, they are overayed over the image
    """

    imageCount = len(images)
    if imagesInRow is 0:
        imagesInRow = imageCount
    if imageCount < imagesInRow:
        imagesInRow = imageCount
    columns = np.ceil(imageCount / imagesInRow)

    fig = plt.figure(figsize=(20, 20))
    fig.subplots_adjust(
        left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)

    for i in range(imageCount):
        axis = fig.add_subplot(columns, imagesInRow, i+1, xticks=[], yticks=[])
        if predictions is not None:
            showImage(images[i], predictions[i], axis=axis)
        else:
            showImage(images[i], axis=axis)

    plt.show()


def plotHistory(*networks, histOnly=False, limits=[1e-3, 1e-2]):
    """
    plot the training history of one or more neural network

    a network is given as tuple [<name>, <network>]
    """

    for network in networks:
        name = network[0]
        if histOnly:
            history = network[1]
        else:
            history = network[1].network.train_history_

        train_loss = np.array([i["train_loss"] for i in history])
        valid_loss = np.array([i["valid_loss"] for i in history])
        plt.plot(train_loss, linewidth=1, label=name + " train")
        plt.plot(valid_loss, linewidth=1, label=name + " valid")
        plt.grid()
        plt.legend()
        plt.xlabel("epoch")
        plt.ylabel("loss")
        # plt.ylim(1e-3, 1e-2)
        if limits:
            plt.ylim(*limits)
        plt.yscale("log")

    plt.show()