self-driving-convnet/preprocess_train_data.py at master · tanelp/self-driving-convnet · 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
from keras.preprocessing.image import load_img, img_to_array
import pandas as pd
import numpy as np
from skimage.exposure import rescale_intensity
from matplotlib.colors import rgb_to_hsv
from config import DataConfig

def make_hsv_data(path):
    df = pd.read_csv(path)
    num_rows = df.shape[0]

    X = np.zeros((num_rows, row, col, 3), dtype=np.uint8)
    for i in range(num_rows):
        if i % 1000 == 0:
            print "Processed " + str(i) + " images..."

        path = df['fullpath'].iloc[i]
        img = load_img(data_path + path, target_size=(row, col))
        img = img_to_array(img)
        img = rgb_to_hsv(img)
        img = np.array(img, dtype=np.uint8)

        X[i] = img

    return X, np.array(df["angle"])

def make_color_data(path):
    df = pd.read_csv(path)
    num_rows = df.shape[0]

    X = np.zeros((num_rows, row, col, 3), dtype=np.uint8)
    for i in range(num_rows):
        if i % 1000 == 0:
            print "Processed " + str(i) + " images..."

        path = df['fullpath'].iloc[i]
        img = load_img(data_path + path, target_size=(row, col))
        img = img_to_array(img)
        img = np.array(img, dtype=np.uint8)

        X[i] = img

    return X, np.array(df["angle"])

def make_grayscale_diff_data(path, num_channels=2):
    df = pd.read_csv(path)
    num_rows = df.shape[0]

    X = np.zeros((num_rows - num_channels, row, col, num_channels), dtype=np.uint8)
    for i in range(num_channels, num_rows):
        if i % 1000 == 0:
            print "Processed " + str(i) + " images..."
        for j in range(num_channels):
            path0 = df['fullpath'].iloc[i - j - 1]
            path1 = df['fullpath'].iloc[i - j]
            img0 = load_img(data_path + path0, grayscale=True, target_size=(row, col))
            img1 = load_img(data_path + path1, grayscale=True, target_size=(row, col))
            img0 = img_to_array(img0)
            img1 = img_to_array(img1)
            img = img1 - img0
            img = rescale_intensity(img, in_range=(-255, 255), out_range=(0, 255))
            img = np.array(img, dtype=np.uint8)

            X[i - num_channels, :, :, j] = img[:, :, 0]
    return X, np.array(df["angle"].iloc[num_channels:])

def make_grayscale_diff_tx_data(path, num_channels=2):
    df = pd.read_csv(path)
    num_rows = df.shape[0]

    X = np.zeros((num_rows - num_channels, row, col, num_channels), dtype=np.uint8)
    for i in range(num_channels, num_rows):
        if i % 1000 == 0:
            print "Processed " + str(i) + " images..."
        path1 = df['fullpath'].iloc[i]
        img1 = load_img(data_path + path1, grayscale=True, target_size=(row, col))
        img1 = img_to_array(img1)
        for j in range(1, num_channels + 1):
            path0 = df['fullpath'].iloc[i - j]
            img0 = load_img(data_path + path0, grayscale=True, target_size=(row, col))
            img0 = img_to_array(img0)

            img = img1 - img0
            img = rescale_intensity(img, in_range=(-255, 255), out_range=(0, 255))
            img = np.array(img, dtype=np.uint8)

            X[i - num_channels, :, :, j - 1] = img[:, :, 0]
    return X, np.array(df["angle"].iloc[num_channels:])

def make_hsv_grayscale_diff_data(path, num_channels=2):
    df = pd.read_csv(path)
    num_rows = df.shape[0]

    X = np.zeros((num_rows - num_channels, row, col, num_channels), dtype=np.uint8)
    for i in range(num_channels, num_rows):
        if i % 1000 == 0:
            print "Processed " + str(i) + " images..."
        for j in range(num_channels):
            path0 = df['fullpath'].iloc[i - j - 1]
            path1 = df['fullpath'].iloc[i - j]
            img0 = load_img(data_path + path0, target_size=(row, col))
            img1 = load_img(data_path + path1, target_size=(row, col))
            img0 = img_to_array(img0)
            img1 = img_to_array(img1)
            img0 = rgb_to_hsv(img0)
            img1 = rgb_to_hsv(img1)
            img = img1[:, :, 2] - img0[:, :, 2]
            img = rescale_intensity(img, in_range=(-255, 255), out_range=(0, 255))
            img = np.array(img, dtype=np.uint8)

            X[i - num_channels, :, :, j] = img
    return X, np.array(df["angle"].iloc[num_channels:])

if __name__ == "__main__":
    config = DataConfig()
    data_path = config.data_path
    row, col = config.height, config.width

    print "Pre-processing phase 1 data..."
    X_train, y_train = make_hsv_grayscale_diff_data("data/train_round1.txt", 4)
    np.save("{}/X_train_round1_hsv_gray_diff_ch4".format(data_path), X_train)
    np.save("{}/y_train_round1_hsv_gray_diff_ch4".format(data_path), y_train)

    X_val, y_val = make_hsv_grayscale_diff_data("data/val_round1.txt", 4)
    np.save("{}/X_train_round1_hsv_gray_diff_ch4".format(data_path), X_val)
    np.save("{}/y_train_round1_hsv_gray_diff_ch4".format(data_path), y_val)

    print "Pre-processing phase 2 data..."
    for i in range(1, 6):
        X_train, y_train = make_hsv_grayscale_diff_data("data/train_round2_part" + str(i) + ".txt", 4)
        np.save("{}/X_train_round2_hsv_gray_diff_ch4_part{}".format(data_path, i), X_train)
        np.save("{}/y_train_round2_hsv_gray_diff_ch4_part{}".format(data_path, i), y_train)