surreal/dataloader.py at master · emilysun201309/surreal · 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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
import os
import numpy as np
import pandas as pd
import torch
from matplotlib import pyplot as plt
from scipy.misc import imread, imresize
from torch.utils.data import Dataset, DataLoader
from scipy.io import loadmat
import csv
import json
from PIL import Image
import tarfile
import glob
import h5py


FOLDER_DATASET = "/out"

class MotionData(Dataset):
    #__depth = []
    #__flow = []
    #__segm = []
    #__normal = []
    __data = []
    __annotation = []
    __img = []
    __keypoint = []

    def __init__(self, folder_dataset, transform=None):
        self.transform = transform
        # Open and load text file including the whole training data
        with open('files.csv', mode='r') as csv_file:
            csv_reader = csv.DictReader(csv_file)
            for row in csv_reader:
                # depth path
                #self.__depth.append(row['depth'])
                # flow path
                #self.__flow.append(row['flow'])
                # segm path
                #self.__segm.append(row['segm'])
                # normal path
                #self.__normal.append(row['normal'])
                self.__data.append(row['data'])
                # annotation path
                self.__annotation.append(row['annotation'])
                #image path
                self.__img.append(row['img'])
                self.__keypoint.append(row['keypoint'])

    # Override to give PyTorch access to any image on the dataset
    def __getitem__(self, index):
        print('load file')
        h5f = h5py.File(self.__data[index],'r')
        print('finish loading')
        data_len = 130
        depth= np.zeros((data_len,240,320))
        try:
            print('load depth')
            depth_temp = h5f['depth'][:]
            print('finish loading')
            length = len(depth_temp)
            #if(length <= data_len):
            #    depth = np.pad(depth,((0,data_len-length),(0,0),(0,0)),'constant')
            depth[:min(data_len,length)] = depth_temp[:min(data_len,length)]
        except:
            print("problem with %d depth data"%index)

        flow = np.zeros((data_len,240,320,2))
        try:
            flow_temp = h5f['gtflow'][:]
            length = len(flow_temp)
            #if(length <= data_len):
            #    flow = np.pad(flow,((0,data_len-length),(0,0),(0,0)),'constant')
            flow[:min(data_len,length)] = flow_temp[:min(data_len,length)]
        except:
            print("problem with %d flow data"%index)
        segm = np.zeros((data_len,240,320))
        try:
            segm_temp = h5f['segm'][:]
            length = len(segm_temp)
            #if(length <= data_len):
            #    segm = np.pad(segm,((0,data_len-length),(0,0),(0,0)),'constant')
            segm[:min(data_len,length)] = segm_temp[:min(data_len,length)]
        except:
            print("problem with %d segm data"%index)
        normal = np.zeros((data_len,240,320,3))
        try:
            normal_temp = h5f['normal'][:]
            length = len(normal_temp)
            #if(length <= data_len):
            #    normal = np.pad(normal,((0,data_len-length),(0,0),(0,0)),'constant')
            normal[:min(data_len,length)] = normal_temp[:min(data_len,length)]
        except:
            print("problem with %d normal data"%index)
        #annotation
        try:
            with open(self.__annotation[index]) as f:
                    annotation = json.load(f)
        except:
            print("problem with annotation")
        #images
        img = np.zeros((data_len,240,320,3),dtype=int)
        try:

            #dimension (H,W,C)

            tarfile.open(self.__img[index]).extractall(self.__img[index].split(".")[0])

            length = len(glob.glob("%s/*/*" %(self.__img[index].split(".")[0])))
            print('image',length)
            for i in range(min(data_len,length)):
                path = glob.glob("%s/*/Image%04d.png" %(self.__img[index].split(".")[0],i))
                print('path',path)
                with Image.open(path[0]) as img_temp:
                    img_temp = img_temp.convert('RGB')
                    img_temp = np.asarray(img_temp)
                    img_temp = img_temp.astype(int)
                    img[i] = img_temp
                #img[i] = torch.from_numpy(img_temp,dtype=torch.int)
        except:
            print("problem with %d image"%index)

        #keypoint
        h5f_keypoint = h5py.File(self.__keypoint[index],'r')

        keypoint = np.zeros((2,43,data_len))
        try:
            keypoint_temp = h5f_keypoint['joints2D'][:]
            print('keypoint shape',keypoint_temp.shape)
            length = keypoint_temp.shape[2]
            #if(length <= data_len):
            #    normal = np.pad(normal,((0,data_len-length),(0,0),(0,0)),'constant')
            keypoint[:,:,:min(data_len,length)] = keypoint_temp[:,:,:min(data_len,length)]
        except:
            print("problem with %d keypoint data"%index)

        # Convert image and label to torch tensors
        depth = torch.from_numpy(np.asarray(depth))


        flow = torch.from_numpy(np.asarray(flow))

        segm = torch.from_numpy(np.asarray(segm))

        normal = torch.from_numpy(np.asarray(normal))

        img = torch.from_numpy(np.asarray(img))

        keypoint = torch.from_numpy(np.asarray(keypoint))

        return depth,flow,segm,normal,annotation,img,keypoint

    # Override to give PyTorch size of dataset
    def __len__(self):
        return len(self.__data)

def main():
    dset_train = MotionData(FOLDER_DATASET)
    train_loader = DataLoader(dset_train, batch_size=1, shuffle=True, num_workers=1)
    depth,flow,segm,normal,annotation,img,keypoint = next(iter(train_loader))
    print("keypoint shape",keypoint.shape)
    print('Batch shape:',depth.numpy().shape, flow.numpy().shape,img.numpy().shape)
    frames = int(img.numpy().shape[1]/10)
    for i in range(frames):
        idx = i * 10
        image = img.numpy()[0,idx,:,:,:]

        ax0 = plt.subplot(231)
        ax0.imshow(image)
        # plt.show()
        #plt.savefig('render/image%d.png'%i)
        #print(image)
        #plt.close()
        joints_loc = keypoint.numpy()[0,:,:,idx]
        print('joints shape',joints_loc.shape)
        ax0.plot(joints_loc[0,:], 240-joints_loc[1,:], 'r+')
        #plt.savefig('image%d.png'%f)


        ax1 = plt.subplot(232)
        ax1.imshow(depth.numpy()[0,idx,:,:])
        #plt.show()
        #plt.waitforbuttonpress()
        #plt.savefig('render/depth%d.png'%i)
        ax2 = plt.subplot(233)
        ax2.imshow(segm.numpy()[0,idx,:,:])
        #plt.show()
        ax3 = plt.subplot(234)
        ax3.imshow(flow.numpy()[0,idx,:,:,0])
        ax4 = plt.subplot(235)
        ax4.imshow(normal.numpy()[0,idx,:,:,:])
        plt.waitforbuttonpress()
        #plt.savefig('render/segm%d.png'%i)
        plt.close()
if __name__ == '__main__':
    main()