lipnet-replication/vsrDataset.py at master · klauscc/lipnet-replication · 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
import os
import numpy as np
import hickle as hkl
import keras.backend as K
from PIL import Image as pil_image

class vsrDatasetLip():

    def __init__(self, target_size=[50,100], re_generate=False ,data_dir='./data/vsr', hkl_path='./data/hkl/vsr.hkl'):
        self.data_root = data_dir
        self.hkl_path = hkl_path
        self.re_generate = re_generate

        self.x_train = None
        self.y_train = None
        self.x_val = None
        self.y_val = None
        self.target_size = target_size


    """
    get image data from data dir
    """
    def get_data_from_dir(self):
        data_path = self.data_root

        channel = 3

        persons = 7
        words = 10
        sample_train = 20
        sample_val = 15
        timespecs = 30

        train_n = (persons-1)*words*sample_train
        val_n = words*sample_val

        x_train = np.zeros([train_n,timespecs, self.target_size[0], self.target_size[1],channel])
        y_train = np.zeros([train_n])

        x_val = np.zeros([val_n,timespecs, self.target_size[0], self.target_size[1],channel])
        y_val = np.zeros([val_n])

        i=0
        for person in xrange(1,persons):
            for word in xrange(1, words+1):
                word_path = os.path.join(data_path,str(person),str(word))
                sample_paths = os.listdir(word_path)
                for sample in sample_paths:
                    image_path = os.path.join(word_path,sample)
                    for t in range(30):
                        path = image_path+'/'+str(t)+'.bmp'
                        print 'train_set,loading',path
                        img = self.load_image(path ,target_size=self.target_size)
                        img = np.asarray(img,dtype=K.floatx())
                        img = self.preprocess_input(img)
                        x_train[i,t,:,:] = img
                        y_train[i] = word-1
                    i = i+1

        person = 7
        i=0
        for word in xrange(1, words+1):
            word_path = os.path.join(data_path,str(person),str(word))
            sample_paths = os.listdir(word_path)
            for sample in sample_paths:
                image_path = os.path.join(word_path,sample)
                for t in range(30):
                    path = image_path+'/'+str(t)+'.bmp'
                    print 'val_set, loading',path
                    img = self.load_image(path ,target_size=self.target_size)
                    img = np.asarray(img,dtype=float)
                    img = self.preprocess_input(img)
                    x_val[i,t,:,:] = img
                    y_val[i] = word-1
                i=i+1

        self.x_train = x_train
        self.y_train = y_train
        self.x_val = x_val
        self.y_val = y_val
        return [x_train,y_train,x_val,y_val]


    def load_data(self):
        hkl_path = self.hkl_path
        if os.path.isfile(hkl_path) and not self.re_generate:
            return hkl.load(hkl_path)
        else:
            z = self.get_data_from_dir()
            hkl_dir = os.path.dirname(hkl_path)
            os.system('mkdir -p ' + hkl_dir)
            hkl.dump(z, hkl_path)
            return z

    def load_image(self,path, grayscale=False, target_size=None):
        img = pil_image.open(path)
        if grayscale:
            img = img.convert('L')
        else:
            img = img.convert('RGB')

        if target_size:
            img = img.resize((target_size[1],target_size[0]))
        return img

    def preprocess_input(self,x):
        x /= 255.
        x -= 0.5
        x *= 2.
        return x


if __name__ == "__main__":
    dataset = DatasetLip()
    dataset.get_data_from_dir()