ClutterObjDetectionFCN/dataset.py at master · sujaymanb/ClutterObjDetectionFCN · 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
import torch
import torch.utils.data as data
import glob
import os
import utils
import numpy as np

# TODO: validation
class SuctionData(data.Dataset):
    def __init__(self, root_dir, mode='train', transform=None):
        """
        args:
            root_dir: root directory of path
            mode: load train or test dataset
            transform: apply any transforms on the data or None
        """
        self.root_dir = root_dir
        self.mode = mode
        self.transform = transform

        # load file list
        self.color_list = glob.glob(os.path.join(root_dir, 'data', self.mode, 'color', '*.png'))
        self.depth_list = glob.glob(os.path.join(root_dir, 'data', self.mode, 'depth', '*.png'))
        self.label_list = glob.glob(os.path.join(root_dir, 'data', self.mode, 'label', '*.png'))

        print(len(self.color_list),len(self.depth_list),len(self.label_list))


    def __len__(self):
        return len(self.color_list)


    def __getitem__(self, idx):
        if torch.is_tensor(idx):
            idx = idx.tolist()

        color = utils.rgb2array(self.color_list[idx],hwc=False)
        color = torch.from_numpy(color)
        depth = utils.depth2array(self.depth_list[idx],hwc=False)
        depth = torch.from_numpy(depth)
        if self.transform:
            color = self.transform(color)
            depth = self.transform(depth)
        if self.mode == 'train':
            label = utils.label2array(self.label_list[idx],hwc=False)
            label = torch.from_numpy(label.astype(np.float32))
            if self.transform:
                label = self.transform(label)
            sample = {'color': color,
                      'depth': depth,
                      'label': label}
        else:
            sample = {'color': color,
                      'depth': depth}

        return sample