FroathOpticalFlow/dataManager.py at master · s70c3/FroathOpticalFlow · 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
import os
import random

import cv2
import numpy as np


class DataManager(object):
    """Class for generating batches of frames from video"""

    def __init__(self, videoPath):
        """
        Constructor
        Parameters:
            :videoPath(str) - path for the input video
        Returns:
        """
        if os.path.exists(videoPath):
            self.video = cv2.VideoCapture(videoPath)
            self.fps = int(self.video.get(cv2.CAP_PROP_FPS))
        else:
            raise FileNotFoundError('No video file: '+videoPath)

    def preprocessing(self, frame):
        """
         Produce input for neural network.
         Parameters:
            :frame (np.ndarray): source image,
         Returns:
            :np.ndarray - contrast gray image
         """
        import cv2
        source = frame.copy()
        frame = cv2.GaussianBlur(frame, (11, 11), 0)
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        gray = cv2.equalizeHist(gray)
        img = cv2.resize(gray, (256, 256))
        # img = img/255.
        img = img.reshape(1, 256, 256)
        return img, gray, source

## TODO: fix for normal pytorch batches
    def get_frames_gen(self, batch_size=32, num=np.inf):
        """
          Batches generator for neural network input.
          Parameters:
              :batch_size(int) - size of the images batch for Unet.
              :num(int) - maximum number of frames.
          """
        counter = 0
        while self.video.isOpened() and counter < num:
            imgs = []
            imgs_gray = []
            sources = []
            for i in range(batch_size):
                ret, img = self.video.read()
                if not ret:
                    break
                img, img_gray, source = self.preprocessing(img)
                imgs.append(img)
                imgs_gray.append(img_gray)
                sources.append(source)
                counter += 1
            yield np.asarray(imgs), np.asarray(imgs_gray), np.asarray(sources)
        self.video.release()


class FolderDataManager(object):
    """Class for generating batches of frames from folder with imgs and gt"""

    def __init__(self, folderPath, img_prefix, mask_prefix):
        """
        Constructor
        Parameters:
            :videoPath(str) - path for the input video
        Returns:
        """
        if os.path.exists(folderPath):
            self.img_list = os.listdir(os.path.join(folderPath, img_prefix))
            self.mask_dir = (os.path.join(folderPath, mask_prefix))
            self.img_dir = (os.path.join(folderPath, img_prefix))
        else:
            raise FileNotFoundError('No folder: '+folderPath)

    def preprocessing(self, frame):
        """
         Produce input for neural network.
         Parameters:
            :frame (np.ndarray): source image,
         Returns:
            :np.ndarray - contrast gray image
         """
        import cv2
        source = frame.copy()
        frame = cv2.GaussianBlur(frame, (11, 11), 0)
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        gray = cv2.equalizeHist(gray)
        img = cv2.resize(gray, (256, 256))
        # img = img/255.
        img = img.reshape(1, 256, 256)
        return img, gray, source

## TODO: fix for normal pytorch batches
    def get_frames_gen(self, batch_size=32, num=None):
        """
          Batches generator for neural network input.
          Parameters:
              :batch_size(int) - size of the images batch for Unet.
              :num(int) - maximum number of frames.
          """
        counter = 0
        if num is not None:
            self.img_list = random.sample(self.img_list, num)
        while counter < len(self.img_list):
            imgs = []
            imgs_gray = []
            sources = []
            masks = []
            for i in range(batch_size):
                if counter> len(self.img_list):
                    break
                img_name = self.img_list[counter]
                img  = cv2.imread(os.path.join(self.img_dir, img_name))
                mask = cv2.imread(os.path.join(self.mask_dir, img_name))
                img, img_gray, source = self.preprocessing(img)
                imgs.append(img)
                imgs_gray.append(img_gray)
                sources.append(source)
                masks.append(mask)
                counter += 1
            yield np.asarray(imgs), np.asarray(imgs_gray), np.asarray(sources), np.asarray(masks)