AtomDetection/utils.py at master · cliulinnaeus/AtomDetection · 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
import sys
sys.path.append('../AtomDetector/')
import numpy as np
import matplotlib.pyplot as plt
import time
import simulator


def visualize(mat2d, figsize=5, title=None, xlabel=None, ylabel=None):

    fig = plt.figure(figsize=(figsize, figsize))

    ax = fig.add_subplot(111)
    if title == None:
        ax.set_title('colorMap')
    else:
        ax.set_title(title)
    plt.imshow(mat2d, cmap='hot')
    ax.set_aspect('equal')
    cax = fig.add_axes([0.12, 0.1, 0.78, 0.8])
    cax.get_xaxis().set_visible(False)
    cax.get_yaxis().set_visible(False)
    cax.patch.set_alpha(0)
    cax.set_frame_on(False)
    cbaxes = fig.add_axes([0.95, 0.1, 0.03, 0.8])
    plt.colorbar(orientation='vertical', cax=cbaxes)
    if xlabel == None or ylabel == None:
        ax.set_xlabel("x [pixels]")
        ax.set_ylabel("y [pixels]")
    else:
        ax.set_xlabel(xlabel)
        ax.set_ylabel(ylabel)

    plt.show()


def gaussian_kernel(img_size, variance, verbose=False):
    x = y = np.linspace(0, img_size-1, img_size)
    xgrid, ygrid = np.meshgrid(x, y)
    def _gaussian(x, y, x0, y0, s):
        return np.exp(-(x-x0)**2 / (2*s)) * np.exp(-(y-y0)**2 / (2*s))
    kernel = _gaussian(xgrid, ygrid, img_size//2, img_size//2, variance) * 1/(2*np.pi*variance)

    if verbose:
        visualize(kernel)
    return kernel

def gaussian(x, y, x0, y0, s):
        return np.exp(-(x-x0)**2 / (2*s)) * np.exp(-(y-y0)**2 / (2*s))


def create_signal(img_size, x0, y0):
    signals = []

    for i in range(x0.shape[0]):
        x = y = np.linspace(0, img_size-1, img_size)
        xgrid, ygrid = np.meshgrid(x, y)
        signal = gaussian(xgrid, ygrid, x0[i], y0[i], 5)
        signals.append(signal)

    signals = np.array(signals)
    return np.sum(signals, axis=0)


def make_circ_mask(shape, radii):
    h, w = shape
    Y, X = np.ogrid[:h, :w]
    center = (int(w / 2), int(h / 2))
    dist_from_center = np.sqrt((X - center[0])**2 + (Y - center[1])**2)

    mask = dist_from_center <= radii
    mask = np.where(mask, 1., 0.)
    return mask


def make_square_mask(shape, height, width):
    mask = np.zeros(shape)
    H, W = shape
    x = (H - height) // 2
    y = (W - width) // 2
    mask[x:x+height+1, y:y+width+1] = 1
    return mask


def fftconvolve_wrap(img, kernel):
    # padded_img = np.zeros(img.shape * 3)
    # padded_img
    padded_img = np.pad(img, img.shape[0], mode='wrap')
    padded_kernel = np.pad(kernel, kernel.shape[0], mode='constant')
    padded_img_fourier = np.fft.fft2(padded_img)
    padded_kernel_fourier = np.fft.fft2(padded_kernel)
    padded_result = np.abs(np.fft.fftshift(np.fft.ifft2(padded_img_fourier * padded_kernel_fourier)))
    H, W = img.shape
    result = padded_result[H:2*H, W:2*W]
    return result


# data label pairs generator
# SNR must be an array
def generate_data_labels(img_size, x0, y0, SNR, N):
    prev_time = time.time()

    sim = simulator.simulator(img_size, 1)

    data_tr_for_different_snr = []
    labels_tr_for_different_snr = []
    for snr in SNR:
        data_tr = []
        labels_tr = []
        for i in range(N):
            r = np.random.rand()
            if r > 0.5:
                labels_tr.append(1)

                data = sim.create_simulation_from_SNR(x0, y0, snr)
                data_tr.append(data)
            else:
                labels_tr.append(0)
                data = sim.create_simulation_from_SNR(x0, y0, snr, no_atom=True)
                data_tr.append(data)
        data_tr_for_different_snr.append(np.array(data_tr))
        labels_tr_for_different_snr.append(np.array(labels_tr))

    print(f"time used: {time.time() - prev_time}")
    return np.array(data_tr_for_different_snr), np.array(labels_tr_for_different_snr)


# normalize to have a variance of 1, mean of 0
def normalize(data_batch):
    mean = np.mean(data_batch, axis=0)
    std = np.std(data_batch, axis=0)

    return (data_batch - mean) / std