Python-Deep-Learning-Projects/Chapter12/crop_resize_transform.py at master · learner-python-R/Python-Deep-Learning-Projects · 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
"""This module contains functions to crop and resize images."""

import os
import cv2 as cv
import numpy as np
import pandas as pd


def image_cropping(image_id, joints, crop_pad_inf=1.4, crop_pad_sup=1.6,
                   shift=5, min_dim=100):
    """Crop Function."""
    # # image cropping
    # load the image
    image = cv.imread('FLIC-full/images/%s' % (image_id))
    # convert joint list to array
    joints = np.asarray([int(float(p)) for p in joints])
    # reshape joints to shape (7*2)
    joints = joints.reshape((len(joints) // 2, 2))
    # transform joints to list of (x,y) tuples
    posi_joints = [(j[0], j[1]) for j in joints if j[0] > 0 and j[1] > 0]
    # obtain the bounding rectangle using opencv boundingRect
    x_loc, y_loc, width, height = cv.boundingRect(np.asarray([posi_joints]))
    if width < min_dim:
        width = min_dim
    if height < min_dim:
        height = min_dim

    # # bounding rect extending
    inf, sup = crop_pad_inf, crop_pad_sup
    r = sup - inf
    # define width padding
    pad_w_r = np.random.rand() * r + inf  # inf~sup
    # define height padding
    pad_h_r = np.random.rand() * r + inf  # inf~sup
    # adjust x, y, w and h by the defined padding
    x_loc -= (width * pad_w_r - width) / 2
    y_loc -= (height * pad_h_r - height) / 2
    width *= pad_w_r
    height *= pad_h_r

    # # shifting
    x_loc += np.random.rand() * shift * 2 - shift
    y_loc += np.random.rand() * shift * 2 - shift

    # # clipping
    x_loc, y_loc, width, height = [int(z) for z in [x_loc, y_loc,
                                                    width, height]]
    x_loc = np.clip(x_loc, 0, image.shape[1] - 1)
    y_loc = np.clip(y_loc, 0, image.shape[0] - 1)
    width = np.clip(width, 1, image.shape[1] - (x_loc + 1))
    height = np.clip(height, 1, image.shape[0] - (y_loc + 1))
    image = image[y_loc: y_loc + height, x_loc: x_loc + width]

    # # joint shifting
    # adjust joint coordinates onto the padded image
    joints = np.asarray([(j[0] - x_loc, j[1] - y_loc) for j in joints])
    joints = joints.flatten()

    return image, joints


def image_resize(image, joints, new_size=224):
    """Resize Function."""
    orig_h, orig_w = image.shape[:2]
    joints[0::2] = joints[0::2] / float(orig_w) * new_size
    joints[1::2] = joints[1::2] / float(orig_h) * new_size
    image = cv.resize(image, (new_size, new_size),
                      interpolation=cv.INTER_NEAREST)
    return image, joints


def model_data(image_ids, joints, train=True):
    """Function to generate train and test data."""
    if train:
        # empty list
        train_img_joints = []
        # create train directory inside FLIC-full
        if not os.path.exists(os.path.join(os.getcwd(), 'FLIC-full/train')):
            os.mkdir('FLIC-full/train')

        for i, (image, joint) in enumerate(zip(image_ids, joints)):
            # crop the image using the joint coordinates
            image, joint = image_cropping(image, joint)
            # resize the cropped image to shape (224*224*3)
            image, joint = image_resize(image, joint)
            # save the image in train folder
            cv.imwrite('FLIC-full/train/train{}.jpg'.format(i), image)
            # store joints and image id/file name of the saved image in
            # the initialized list
            train_img_joints.append(['train{}.jpg'.format(i)] + joint.tolist())

        # convert to a dataframe and save as a csv
        pd.DataFrame(train_img_joints
                     ).to_csv('FLIC-full/train/train_joints.csv',
                              index=False, header=False)
    else:
        # empty list
        test_img_joints = []
        # create test directory inside FLIC-full
        if not os.path.exists(os.path.join(os.getcwd(), 'FLIC-full/test')):
            os.mkdir('FLIC-full/test')

        for i, (image, joint) in enumerate(zip(image_ids, joints)):
            # crop the image using the joint coordinates
            image, joint = image_cropping(image, joint)
            # resize the cropped image to shape (224*224*3)
            image, joint = image_resize(image, joint)
            # save the image in test folder
            cv.imwrite('FLIC-full/test/test{}.jpg'.format(i), image)
            # store joints and image id/file name of the saved image
            # in the initialized list
            test_img_joints.append(['test{}.jpg'.format(i)] + joint.tolist())

        # convert to a dataframe and save as a csv
        pd.DataFrame(test_img_joints).to_csv('FLIC-full/test/test_joints.csv',
                                             index=False, header=False)