ClothesClassifier/utils.py at main · Nerocraft4/ClothesClassifier · 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
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
import numpy as np

def rgb2gray(rgb):
    return np.dot(rgb[..., :3], [0.2989, 0.5870, 0.1140])


def Sigmoid(s, t, b):
    return 1.0 / (1.0 + np.exp(-np.double(b) * (np.double(s) - np.double(t))))


def TripleSigmoid_E(s, tx, ty, alfa_x, alfa_y, bx, by, be, ex, ey, angle_e):
    sT = np.double(s.T) - np.hstack([tx, ty])
    sR = np.hstack([sT[:, 0].reshape((-1, 1)) * np.cos(alfa_y) + sT[:, 1].reshape((-1, 1)) * np.sin(alfa_y),
                    -sT[:, 0].reshape((-1, 1)) * np.sin(alfa_x) + sT[:, 1].reshape((-1, 1)) * np.cos(alfa_x)])
    sRE = np.hstack([sT[:, 0].reshape((-1, 1)) * np.cos(angle_e) + sT[:, 1].reshape((-1, 1)) * np.sin(angle_e),
                     -sT[:, 0].reshape((-1, 1)) * np.sin(angle_e) + sT[:, 1].reshape((-1, 1)) * np.cos(angle_e)])
    ex = (ex == 0.0) + ex
    ey = (ey == 0.0) + ey

    y = 1.0 / np.hstack([1.0 + np.exp(-sR * np.hstack([by, bx])),
                         1.0 + np.exp(
                             -be * (np.sum((sRE / np.hstack([ex, ey])) ** 2.0, axis=1).reshape((-1, 1)) - 1.0))])

    return np.prod(y, axis=1).reshape((-1, 1))


def RGB2Lab(Ima):
    # RGB > XYZ transformation matrix (sRGB with D65)
    M = np.vstack(([0.412424, 0.357579, 0.180464], [0.212656, 0.715158, 0.0721856], [0.0193324, 0.119193, 0.950444]))
    Xn = 0.9505
    Yn = 1.0000
    Zn = 1.0891

    Ima = Ima / 255.0
    ImaLab = np.zeros_like(Ima, dtype=float)

    fRGB = np.reshape(Ima, (-1, 3)).T
    lRGB = (fRGB <= 0.04045) * (fRGB / 12.92) + (fRGB > 0.04045) * (((fRGB + 0.055) / 1.055) ** 2.4)
    XYZ = np.dot(M, lRGB)

    f_X2 = (XYZ[0] / Xn > 0.008856) * ((XYZ[0] / Xn) ** (1.0 / 3.0)) + (XYZ[0] / Xn <= 0.008856) * (
            7.787 * (XYZ[0] / Xn) + (16.0 / 116.0))
    f_Y2 = (XYZ[1] / Yn > 0.008856) * ((XYZ[1] / Yn) ** (1.0 / 3.0)) + (XYZ[1] / Yn <= 0.008856) * (
            7.787 * (XYZ[1] / Yn) + (16.0 / 116.0))
    f_Z2 = (XYZ[2] / Zn > 0.008856) * ((XYZ[2] / Zn) ** (1.0 / 3.0)) + (XYZ[2] / Zn <= 0.008856) * (
            7.787 * (XYZ[2] / Zn) + (16.0 / 116.0))

    L2 = (XYZ[1] / Yn > 0.008856) * ((116.0 * ((XYZ[1] / Yn) ** (1.0 / 3.0))) - 16.0) + (XYZ[1] / Yn <= 0.008856) * (
            903.3 * (XYZ[1] / Yn))
    a2 = 500.0 * (f_X2 - f_Y2)
    b2 = 200.0 * (f_Y2 - f_Z2)

    ImaLab = np.reshape(np.vstack((L2.flatten(), a2.flatten(), b2.flatten())).T, ImaLab.shape)

    return ImaLab


def get_color_prob(ima, positions=None, patchSize=1):
    # Constants
    numColors = 11  # Number of colors
    numAchromatics = 3  # Number of achromatic colors
    numChromatics = numColors - numAchromatics  # Number of chromatic colors

    # Initializations
    numLevels = np.size(thrL) - 1  # Number of Lightness levels in the model

    # Image conversion: sRGB to CIELab
    Lab = RGB2Lab(ima)
    if positions is not None:
        if patchSize == 1:
            Lab = Lab[positions[:, 0], positions[:, 1], :].reshape((1, -1, 3))
        else:
            LabPatch = np.zeros((positions.shape[0], (2 * np.trunc(patchSize / 2) + 1) ** 2, 3))
            padSz = (int(np.trunc(patchSize / 2)), int(np.trunc(patchSize / 2)))
            Lab = np.pad(Lab, (padSz, padSz, (0, 0)), 'symmetric')
            positions += padSz[0]
            c = 0
            for x in range(-padSz[0], padSz[0] + 1):
                for y in range(-padSz[0], padSz[0] + 1):
                    LabPatch[:, c, :] = Lab[positions[:, 0] + y, positions[:, 1] + x, :]
                    c += 1
            Lab = LabPatch

    S = np.shape(Lab)
    if Lab.ndim == 2:
        L = Lab[:, 0].flatten()
        a = Lab[:, 1].flatten()
        b = Lab[:, 2].flatten()
        nr = S[0]
        nc = 1  # Image dimensions: rows, columns, and channels
    else:
        L = Lab[:, :, 0].flatten()
        a = Lab[:, :, 1].flatten()
        b = Lab[:, :, 2].flatten()
        nr = S[0]
        nc = S[1]  # Image dimensions: rows, columns, and channels

    npix = nr * nc  # Number of pixels
    CD = np.zeros((npix, numColors))  # Color descriptor to store results

    # Assignment of the sample to its corresponding level
    m = np.zeros(np.shape(L))
    m[np.where(L == 0)[0]] = 1  # Pixels with L=0 assigned to level 1
    for k in range(1, numLevels + 1):
        m = m + np.double(thrL[k - 1] < L) * np.double(L <= thrL[k]) * np.double(k)

    m = m.astype(int) - 1

    # Computing membership values to chromatic categories
    for k in range(numChromatics):
        tx = np.reshape(parameters[k, 0, m], (npix, 1))
        ty = np.reshape(parameters[k, 1, m], (npix, 1))
        alfa_x = np.reshape(parameters[k, 2, m], (npix, 1))
        alfa_y = np.reshape(parameters[k, 3, m], (npix, 1))
        beta_x = np.reshape(parameters[k, 4, m], (npix, 1))
        beta_y = np.reshape(parameters[k, 5, m], (npix, 1))
        beta_e = np.reshape(parameters[k, 6, m], (npix, 1))
        ex = np.reshape(parameters[k, 7, m], (npix, 1))
        ey = np.reshape(parameters[k, 8, m], (npix, 1))
        angle_e = np.reshape(parameters[k, 9, m], (npix, 1))  # figure;plot(angle_e); show()
        CD[:, k] = (np.double(beta_e != 0.0) * TripleSigmoid_E(np.vstack((a, b)), tx, ty, alfa_x, alfa_y, beta_x,
                                                               beta_y, beta_e, ex, ey, angle_e)).T

    # Computing membership values to achromatic categories
    valueAchro = np.squeeze(np.maximum(1.0 - np.reshape(np.sum(CD, axis=1), (npix, 1)), np.zeros((npix, 1))))
    CD[:, numChromatics + 0] = valueAchro * Sigmoid(L, paramsAchro[0, 0], paramsAchro[0, 1])
    CD[:, numChromatics + 1] = valueAchro * Sigmoid(L, paramsAchro[1, 0], paramsAchro[1, 1]) * Sigmoid(L, paramsAchro[
        2, 0], paramsAchro[2, 1])
    CD[:, numChromatics + 2] = valueAchro * Sigmoid(L, paramsAchro[3, 0], paramsAchro[3, 1])

    # Color descriptor with color memberships to all the categories (one color in each plane)
    if positions is None or patchSize > 1:
        CD = np.reshape(CD, (nr, nc, numColors))
    if patchSize > 1:
        CD = np.sum(CD, axis=1)
        CD = CD / np.tile(np.sum(CD, axis=1).reshape(-1, 1), (1, numColors))

    if Lab.ndim == 2:
        CD = np.reshape(CD, (-1, CD.shape[2]))

    CD = CD / np.expand_dims(np.sum(CD, axis=len(CD.shape) - 1), axis=len(CD.shape) - 1)
    return CD


colors = np.array(['Red', 'Orange', 'Brown', 'Yellow', 'Green', 'Blue', 'Purple', 'Pink', 'Black', 'Grey', 'White'])

parameters = np.array(
    [[[4.24199636e-01, 2.29563220e-01, -1.17335858e-01, -4.43169080e-01, 0.00000000e+00, 0.00000000e+00],
      [2.49359126e-01, 6.63190063e-01, 5.18007668e-01, 1.07591884e+00, 0.00000000e+00, 0.00000000e+00],
      [-3.90823298e-02, 3.85324254e-02, 2.36880030e-01, 4.62878817e-01, 0.00000000e+00, 0.00000000e+00],
      [-9.86970116e-01, -8.51969023e-01, -7.94921771e-01, -9.95040273e-01, 0.00000000e+00, 0.00000000e+00],
      [9.00991049e-01, 5.24362727e-01, 9.99376086e-01, 9.39107161e-01, 0.00000000e+00, 0.00000000e+00],
      [1.72499209e+00, 5.00000000e+00, 5.65279390e-01, 7.54842803e-01, 0.00000000e+00, 0.00000000e+00],
      [9.84405931e+00, 6.03260303e+00, 6.81218149e+00, 7.31858270e+00, -0.00000000e+00, -0.00000000e+00],
      [5.88939523e+00, 6.45705135e+00, 5.38109004e+00, 6.06365916e+00, 0.00000000e+00, 0.00000000e+00],
      [7.47432124e+00, 7.87254713e+00, 6.98237582e+00, 7.50204671e+00, 0.00000000e+00, 0.00000000e+00],
      [4.05071773e-02, 3.07016320e-01, 3.41770788e-01, 4.28449446e-01, 0.00000000e+00, 0.00000000e+00]],
     [[0.00000000e+00, 0.00000000e+00, -1.17335858e-01, -4.43169080e-01, -5.65458039e-01, -1.25643027e+00],
      [0.00000000e+00, 0.00000000e+00, 5.18007668e-01, 1.07591884e+00, 1.16327519e+00, 1.81438090e+00],
      [0.00000000e+00, 0.00000000e+00, 7.75874556e-01, 5.75756054e-01, 4.49437961e-01, 4.49316962e-01],
      [0.00000000e+00, 0.00000000e+00, -5.02014629e-01, -1.72492737e-01, -2.76553982e-01, -3.06421502e-01],
      [0.00000000e+00, 0.00000000e+00, 5.65279390e-01, 7.54842803e-01, 1.99511862e+00, 1.02801414e+00],
      [0.00000000e+00, 0.00000000e+00, 5.17227023e-01, 4.78353741e-01, 8.39910145e-01, 7.90983657e-01],
      [-0.00000000e+00, -0.00000000e+00, 6.81218149e+00, 7.31858270e+00, 1.00000000e+02, 1.00000000e+02],
      [0.00000000e+00, 0.00000000e+00, 5.38109004e+00, 6.06365916e+00, 5.36820773e+00, 6.03989989e+00],
      [0.00000000e+00, 0.00000000e+00, 6.98237582e+00, 7.50204671e+00, 6.90048122e+00, 7.39318578e+00],
      [0.00000000e+00, 0.00000000e+00, 3.41770788e-01, 4.28449446e-01, 4.32023987e-01, -2.08201846e-02]],
     [[4.24199636e-01, 2.29563220e-01, -1.17335858e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [2.49359126e-01, 6.63190063e-01, 5.18007668e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [5.83826210e-01, 7.18827304e-01, 1.06878170e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [2.54038810e-01, 1.19930098e-01, 1.16144803e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [1.72499209e+00, 5.00000000e+00, 5.17227023e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [8.44028374e-01, 6.92404886e-01, 8.44786041e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [9.84405931e+00, 6.03260303e+00, 6.81218149e+00, -0.00000000e+00, -0.00000000e+00, -0.00000000e+00],
      [5.88939523e+00, 6.45705135e+00, 5.38109004e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [7.47432124e+00, 7.87254713e+00, 6.98237582e+00, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
      [4.05071773e-02, 3.07016320e-01, 3.41770788e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00]],
     [[0.00000000e+00, 0.00000000e+00, 0.00000000e+00, -4.43169080e-01, -5.65458039e-01, -1.25643027e+00],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.07591884e+00, 1.16327519e+00, 1.81438090e+00],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 1.39830359e+00, 1.29424234e+00, 1.26437482e+00],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 9.90182597e-02, 2.14168287e-01, 2.83461782e-01],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 4.78353741e-01, 8.39910145e-01, 7.90983657e-01],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 7.28741639e-01, 8.56004138e-01, 9.57414544e-01],
      [-0.00000000e+00, -0.00000000e+00, -0.00000000e+00, 7.31858270e+00, 1.00000000e+02, 1.00000000e+02],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 6.06365916e+00, 5.36820773e+00, 6.03989989e+00],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 7.50204671e+00, 6.90048122e+00, 7.39318578e+00],
      [0.00000000e+00, 0.00000000e+00, 0.00000000e+00, 4.28449446e-01, 4.32023987e-01, -2.08201846e-02]],
     [[4.24199636e-01, 2.29563220e-01, -1.17335858e-01, -4.43169080e-01, -5.65458039e-01, -1.25643027e+00],
      [2.49359126e-01, 6.63190063e-01, 5.18007668e-01, 1.07591884e+00, 1.16327519e+00, 1.81438090e+00],
      [1.82483514e+00, 1.69072643e+00, 1.68694113e+00, 1.66981459e+00, 1.78496461e+00, 1.85425811e+00],
      [2.34898971e+00, 2.10250398e+00, 1.90897812e+00, 1.89047978e+00, 1.72030985e+00, 1.74612759e+00],
      [8.44028374e-01, 6.92404886e-01, 8.44786041e-01, 7.28741639e-01, 8.56004138e-01, 9.57414544e-01],
      [1.95218383e+00, 9.58110786e-01, 5.99760697e-01, 6.43832855e-01, 7.36644526e-01, 9.00009936e-01],
      [9.84405931e+00, 6.03260303e+00, 6.81218149e+00, 7.31858270e+00, 1.00000000e+02, 1.00000000e+02],
      [5.88939523e+00, 6.45705135e+00, 5.38109004e+00, 6.06365916e+00, 5.36820773e+00, 6.03989989e+00],
      [7.47432124e+00, 7.87254713e+00, 6.98237582e+00, 7.50204671e+00, 6.90048122e+00, 7.39318578e+00],
      [4.05071773e-02, 3.07016320e-01, 3.41770788e-01, 4.28449446e-01, 4.32023987e-01, -2.08201846e-02]],
     [[4.24199636e-01, 2.29563220e-01, -1.17335858e-01, -4.43169080e-01, -5.65458039e-01, -1.25643027e+00],
      [2.49359126e-01, 6.63190063e-01, 5.18007668e-01, 1.07591884e+00, 1.16327519e+00, 1.81438090e+00],
      [3.91978603e+00, 3.67330031e+00, 3.47977445e+00, 3.46127611e+00, 3.29110617e+00, 3.31692392e+00],
      [-2.56831073e+00, -2.59153227e+00, -2.58725114e+00, -2.59660745e+00, -2.63248600e+00, -2.60808914e+00],
      [1.95218383e+00, 9.58110786e-01, 5.99760697e-01, 6.43832855e-01, 7.36644526e-01, 9.00009936e-01],
      [1.01425447e+00, 9.16607781e-01, 8.00824361e-01, 7.55213599e-01, 4.74809095e-01, 5.98984911e-01],
      [9.84405931e+00, 6.03260303e+00, 6.81218149e+00, 7.31858270e+00, 1.00000000e+02, 1.00000000e+02],
      [5.88939523e+00, 6.45705135e+00, 5.38109004e+00, 6.06365916e+00, 5.36820773e+00, 6.03989989e+00],
      [7.47432124e+00, 7.87254713e+00, 6.98237582e+00, 7.50204671e+00, 6.90048122e+00, 7.39318578e+00],
      [4.05071773e-02, 3.07016320e-01, 3.41770788e-01, 4.28449446e-01, 4.32023987e-01, -2.08201846e-02]],
     [[4.24199636e-01, 2.29563220e-01, -1.17335858e-01, -4.43169080e-01, -5.65458039e-01, -1.25643027e+00],
      [2.49359126e-01, 6.63190063e-01, 5.18007668e-01, 1.07591884e+00, 1.16327519e+00, 1.81438090e+00],
      [-9.97514408e-01, -1.02073594e+00, -1.01645481e+00, -1.02581112e+00, -1.06168967e+00, -1.03729281e+00],
      [-1.60987866e+00, -1.84515592e+00, -1.96583055e+00, -2.16285442e+00, -2.13886579e+00, -2.13569227e+00],
      [1.01425447e+00, 9.16607781e-01, 8.00824361e-01, 7.55213599e-01, 4.74809095e-01, 5.98984911e-01],
      [9.00991049e-01, 1.10031784e+00, 6.23673475e-01, 5.00000000e+00, 1.73558321e+00, 1.93131386e+00],
      [9.84405931e+00, 6.03260303e+00, 6.81218149e+00, 7.31858270e+00, 1.00000000e+02, 1.00000000e+02],
      [5.88939523e+00, 6.45705135e+00, 5.38109004e+00, 6.06365916e+00, 5.36820773e+00, 6.03989989e+00],
      [7.47432124e+00, 7.87254713e+00, 6.98237582e+00, 7.50204671e+00, 6.90048122e+00, 7.39318578e+00],
      [4.05071773e-02, 3.07016320e-01, 3.41770788e-01, 4.28449446e-01, 4.32023987e-01, -2.08201846e-02]],
     [[0.00000000e+00, 2.29563220e-01, -1.17335858e-01, -4.43169080e-01, -5.65458039e-01, -1.25643027e+00],
      [0.00000000e+00, 6.63190063e-01, 5.18007668e-01, 1.07591884e+00, 1.16327519e+00, 1.81438090e+00],
      [0.00000000e+00, -2.74359589e-01, -3.95034220e-01, -5.92058089e-01, -5.68069463e-01, -5.64895945e-01],
      [0.00000000e+00, -1.53226390e+00, -1.33391630e+00, -1.10791751e+00, -1.12135837e+00, -1.12147936e+00],
      [0.00000000e+00, 1.10031784e+00, 6.23673475e-01, 5.00000000e+00, 1.73558321e+00, 1.93131386e+00],
      [0.00000000e+00, 5.24362727e-01, 9.99376086e-01, 9.39107161e-01, 1.99511862e+00, 1.02801414e+00],
      [-0.00000000e+00, 6.03260303e+00, 6.81218149e+00, 7.31858270e+00, 1.00000000e+02, 1.00000000e+02],
      [0.00000000e+00, 6.45705135e+00, 5.38109004e+00, 6.06365916e+00, 5.36820773e+00, 6.03989989e+00],
      [0.00000000e+00, 7.87254713e+00, 6.98237582e+00, 7.50204671e+00, 6.90048122e+00, 7.39318578e+00],
      [0.00000000e+00, 3.07016320e-01, 3.41770788e-01, 4.28449446e-01, 4.32023987e-01, -2.08201846e-02]]])

paramsAchro = np.array([[28.28252201, -0.71423449],
                        [28.28252201, 0.71423449],
                        [79.64930057, -0.30674052],
                        [79.64930057, 0.30674052]])

thrL = np.array([0, 31, 42, 51, 66, 76, 150], dtype=np.uint8)