File size: 6,342 Bytes
99a05f0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import cv2
import numpy as np
from torch.utils.data import Dataset
from torchvision.transforms import Normalize
from common import constants

def mask_split(img, num_parts):
    if not len(img.shape) == 2:
        img = img[:, :, 0]
    mask = np.zeros((img.shape[0], img.shape[1], num_parts))
    for i in np.unique(img):
        mask[:, :, i] = np.where(img == i, 1., 0.)
    return np.transpose(mask, (2, 0, 1))

class BaseDataset(Dataset):

    def __init__(self, dataset, mode, model_type='smpl', normalize=False):
        self.dataset = dataset
        self.mode = mode

        print(f'Loading dataset: {constants.DATASET_FILES[mode][dataset]} for mode: {mode}')

        self.data = np.load(constants.DATASET_FILES[mode][dataset], allow_pickle=True)

        self.images = self.data['imgname']

        # get 3d contact labels, if available
        try:
            self.contact_labels_3d = self.data['contact_label']
            # make a has_contact_3d numpy array which contains 1 if contact labels are no empty and 0 otherwise
            self.has_contact_3d = np.array([1 if len(x) > 0 else 0 for x in self.contact_labels_3d])
        except KeyError:
            self.has_contact_3d = np.zeros(len(self.images))

        # get 2d polygon contact labels, if available
        try:
            self.polygon_contacts_2d = self.data['polygon_2d_contact']
            self.has_polygon_contact_2d = np.ones(len(self.images))
        except KeyError:
            self.has_polygon_contact_2d = np.zeros(len(self.images))

        # Get camera parameters - only intrinsics for now
        try:
            self.cam_k = self.data['cam_k']
        except KeyError:
            self.cam_k = np.zeros((len(self.images), 3, 3))

        self.sem_masks = self.data['scene_seg']
        self.part_masks = self.data['part_seg']

        # Get gt SMPL parameters, if available
        try:
            self.pose = self.data['pose'].astype(float)
            self.betas = self.data['shape'].astype(float)
            self.transl = self.data['transl'].astype(float)
            if 'has_smpl' in self.data:
                self.has_smpl = self.data['has_smpl']
            else:
                self.has_smpl = np.ones(len(self.images))
                self.is_smplx = np.ones(len(self.images)) if model_type == 'smplx' else np.zeros(len(self.images))
        except KeyError:
            self.has_smpl = np.zeros(len(self.images))
            self.is_smplx = np.zeros(len(self.images))

        if model_type == 'smpl':
            self.n_vertices = 6890
        elif model_type == 'smplx':
            self.n_vertices = 10475
        else:
            raise NotImplementedError

        self.normalize = normalize
        self.normalize_img = Normalize(mean=constants.IMG_NORM_MEAN, std=constants.IMG_NORM_STD)

    def __getitem__(self, index):
        item = {}

        # Load image
        img_path = self.images[index]
        try:
            img = cv2.imread(img_path)
            img_h, img_w, _ = img.shape
            img = cv2.resize(img, (256, 256), cv2.INTER_CUBIC)
            img = img.transpose(2, 0, 1) / 255.0
        except:
            print('Img: ', img_path)

        img_scale_factor = np.array([256 / img_w, 256 / img_h])

        # Get SMPL parameters, if available
        if self.has_smpl[index]:
            pose = self.pose[index].copy()
            betas = self.betas[index].copy()
            transl = self.transl[index].copy()
        else:
            pose = np.zeros(72)
            betas = np.zeros(10)
            transl = np.zeros(3)

        # Load vertex_contact
        if self.has_contact_3d[index]:
            contact_label_3d = self.contact_labels_3d[index]
        else:
            contact_label_3d = np.zeros(self.n_vertices)

        sem_mask_path = self.sem_masks[index]
        try:
            sem_mask = cv2.imread(sem_mask_path)
            sem_mask = cv2.resize(sem_mask, (256, 256), cv2.INTER_CUBIC)
            sem_mask = mask_split(sem_mask, 133)
        except:
            print('Scene seg: ', sem_mask_path)

        try:
            part_mask_path = self.part_masks[index]
            part_mask = cv2.imread(part_mask_path)
            part_mask = cv2.resize(part_mask, (256, 256), cv2.INTER_CUBIC)
            part_mask = mask_split(part_mask, 26)
        except:
            print('Part seg: ', part_mask_path)

        try:
            if self.has_polygon_contact_2d[index]:
                polygon_contact_2d_path = self.polygon_contacts_2d[index]
                polygon_contact_2d = cv2.imread(polygon_contact_2d_path)
                polygon_contact_2d = cv2.resize(polygon_contact_2d, (256, 256), cv2.INTER_NEAREST)
                # binarize the part mask
                polygon_contact_2d = np.where(polygon_contact_2d > 0, 1, 0)
            else:
                polygon_contact_2d = np.zeros((256, 256, 3))
        except:
            print('2D polygon contact: ', polygon_contact_2d_path)

        if self.normalize:
            img = torch.tensor(img, dtype=torch.float32)
            item['img'] = self.normalize_img(img)
        else:
            item['img'] = torch.tensor(img, dtype=torch.float32)

        if self.is_smplx[index]:
            # Add 6 zeros to the end of the pose vector to match with smpl
            pose = np.concatenate((pose, np.zeros(6)))

        item['img_path'] = img_path
        item['pose'] = torch.tensor(pose, dtype=torch.float32)
        item['betas'] = torch.tensor(betas, dtype=torch.float32)
        item['transl'] = torch.tensor(transl, dtype=torch.float32)
        item['cam_k'] = self.cam_k[index]
        item['img_scale_factor'] = torch.tensor(img_scale_factor, dtype=torch.float32)
        item['contact_label_3d'] = torch.tensor(contact_label_3d, dtype=torch.float32)
        item['sem_mask'] = torch.tensor(sem_mask, dtype=torch.float32)
        item['part_mask'] = torch.tensor(part_mask, dtype=torch.float32)
        item['polygon_contact_2d'] = torch.tensor(polygon_contact_2d, dtype=torch.float32)

        item['has_smpl'] = self.has_smpl[index]
        item['is_smplx'] = self.is_smplx[index]
        item['has_contact_3d'] = self.has_contact_3d[index]
        item['has_polygon_contact_2d'] = self.has_polygon_contact_2d[index]

        return item

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