CS236G_Final_Project/preprocessing.py at main · ncomly/CS236G_Final_Project · 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
import face_alignment
from skimage import io, draw
import os
import numpy as np
import matplotlib.pyplot as plt
import time
import tqdm

celebA_dir = 'C:/Users/nacom/Documents/Datasets/CelebA/img_align_celeba/'
dogFace_dir = 'C:/Users/nacom/Documents/Datasets/DogFaceNet/after_4_bis/'

color_map = dict({
    ('jaw', (0,0,.7)),
    ('eyebrow_l', (0.7, 0, 0)),
    ('eyebrow_r', (0.7, 0, 0)),
    ('nose_bridge', (0.5, 0, 0.5)),
    ('nose_base', (0.5, 0, 0.5)),
    ('eye_l', (0, 0.5, 0)),
    ('eye_r', (0, 0.5, 0)),
    ('mouth', (1, 0.7, 0)),
    ('lips', (1, 0.7, 0)),

})

def draw_seg(start, end):
    return draw.line_aa(start[0], start[1], end[0], end[1])

def extract_features(preds):
    features = [
        ('jaw', preds[:17]),
        ('eyebrow_l', preds[17:22]),
        ('eyebrow_r', preds[22:27]),
        ('nose_bridge', preds[27:31]),
        ('nose_base', preds[31:36]),
        ('eye_l', np.vstack((preds[36:42], preds[36]))),
        ('eye_r', np.vstack((preds[42:48], preds[42]))),
        ('mouth', np.vstack((preds[48:61], preds[48]))),
        ('lips', preds[61:])
    ]
    return features

def draw_image(base, features):
    # draw lines
    for f, v in features:
        for i in range(len(v)-1):
            a = v[i].astype(int)
            b = v[i+1].astype(int)
            cc, rr, val = draw_seg(a,b)
            rr = np.clip(rr, 0, 217)
            cc = np.clip(cc, 0, 177)
            base[rr, cc] = val[:, np.newaxis]*color_map[f]*255 + (1-val)[:, np.newaxis]*base[rr,cc]
    return base

def process_single_image(img, landmarks_dir, data_dir, save_csv=True, save_jpg=True):
    # generate output path
    output_path = os.path.join(landmarks_dir, img)

    # save as img if true, else as csv
    if save_csv:
        output_path_csv = output_path.replace('.jpg', '_l.csv')
    if save_jpg:
        output_path_jpg = output_path.replace('.jpg', '_l.jpg')

    # # do not process if already in folder
    # if not ((save_csv and not os.path.isfile(output_path_csv)) or
    #         (save_jpg and not os.path.isfile(output_path_jpg))):
    #     return
    if os.path.isfile(output_path_jpg):
        return

    # read in image
    img_data = io.imread(os.path.join(data_dir, img))
    # get prediction
    preds = fa.get_landmarks(img_data)
    # error checking
    if not preds:
        print('No Face Detected!')
        return
    # remove outter list
    preds = preds[0]

    # if saving as a csv
    # if save_csv and not os.path.isfile(output_path_csv):
    #     np.savetxt(output_path_csv, preds, delimiter=',')


    # if saving as jpg
    # if save_jpg and not os.path.isfile(output_path_jpg):
    # extract facial features
    features = extract_features(preds)

    # white background
    base = np.ones_like(img_data)*255

    # draw the image
    landmarks = draw_image(base, features)

    # save image
    io.imsave(output_path_jpg, landmarks)


def preprocess_celebA(data_dir):

    # make the Landmarks folder (if not already there)
    landmarks_dir = os.path.join(data_dir, 'landmarks')
    if not os.path.isdir(landmarks_dir):
        print(f'Landmarks Directory Does Not Exit\nCreating...')
        os.mkdir(landmarks_dir)
        print(f' - {landmarks_dir}')
    else:
        print(f'Landmarks Directory Found! - {landmarks_dir}')


    # process dataset
    #   - one at a time to allow for failure
    for img in tqdm.tqdm(os.listdir(data_dir)):
        img_path = os.path.join(data_dir,img)
        if os.path.isfile(img_path):
            process_single_image(img, landmarks_dir, data_dir)

def preprocess_dogFace(data_dir, w=178, h=218):
    # make the crops folder (if not already there)
    crops_dir = os.path.join(data_dir, 'crops')
    if not os.path.isdir(crops_dir):
        print(f'Crops Directory Does Not Exit\nCreating...')
        os.mkdir(crops_dir)
        print(f' - {crops_dir}')
    else:
        print(f'Crops Directory Found! - {crops_dir}')

    # center points
    # hardcoding is bad! but easier...
    c_w = (224//2)-1
    c_h = (224//2)-1


    dogs = os.listdir(data_dir)
    for dog in tqdm.tqdm(dogs):
        if dog == 'crops':
            continue
        dog_path = os.path.join(data_dir, dog)
        for img in os.listdir(dog_path):
            # generate and check output path
            output_path = os.path.join(crops_dir, img).replace('.jpg', '_c.jpg')
            if os.path.isfile(output_path):
                continue
            # read and crop data
            img_data = io.imread(os.path.join(dog_path, img))
            img_data = img_data[ c_h - h//2:c_h + h//2,c_w - w//2:c_w + w//2,:]

            # save image
            io.imsave(output_path, img_data)


if __name__ == '__main__':
    # create face aligner
    fa = face_alignment.FaceAlignment(face_alignment.LandmarksType._2D, flip_input=False, device='cuda')
    preprocess_celebA(celebA_dir)
    preprocess_dogFace(dogFace_dir)