-
Notifications
You must be signed in to change notification settings - Fork 5
/
dataset.py
141 lines (108 loc) · 4 KB
/
dataset.py
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
import os
import numpy as np
import cv2
from torch.utils.data import Dataset, DataLoader
import glob2 as gb
from util import *
from torch.multiprocessing import Manager
from configs import get_cfg_defaults
class WarpDataset(Dataset):
"""Speech2Gesture dataset."""
def __init__(self, config_path, mode):
cfg = get_cfg_defaults()
cfg.merge_from_file(config_path)
cfg = cfg.POSE2IMAGE
all_kp_path = sorted(gb.glob(os.path.join(cfg.PATH.kp_base, '*.npy')))[::]
self.img_extension = cfg.PATH.img_extension
self.img_base = cfg.PATH.img_base
self.mode = mode
self.W_bias = cfg.TRAIN.CROP.W_bias
self.H_bias = cfg.TRAIN.CROP.H_bias
self.img_H = cfg.HYPERPARAM.img_H
self.img_W = cfg.HYPERPARAM.img_W
self.scale = cfg.HYPERPARAM.scale
self.limbs = [[0,8,9],[1,2,5],[2,3],[3,4],[5,6],[6,7],range(101,122),range(80,101)]
if mode == "train":
self.kp_path = all_kp_path[256:]
self.len = len(self.kp_path)
elif mode == "val":
self.kp_path = all_kp_path[:256]
self.len = len(self.kp_path)
if cfg.TRAIN.CACHING:
self.cache_dict = Manager().dict()
self.cfg = cfg
def __len__(self):
return self.len
def get_data(self, idx):
kp_path = self.kp_path[idx]
kp = np.load(kp_path)
if kp.shape[1]==137:
kp = pose137_to_pose122(kp).transpose(1,0)
else:
kp = kp
path = kp_path.split("/")[-1]
filename, _ = os.path.splitext(path)
img_path = os.path.join(self.img_base,filename+self.img_extension)
img = cv2.imread(img_path).transpose(2,0,1)
img = img[:,self.H_bias:self.H_bias+self.img_H,self.W_bias:self.W_bias + self.img_W]/255.0*2.0 - 1.0
scale = self.scale
if scale !=1.0:
img = img.transpose(1,2,0)
img = cv2.resize(img,(int(self.img_H/scale),int(self.img_W/scale)))
img = img.transpose(2,0,1)
kp[:,0] -= self.W_bias
kp[:,1] -= self.H_bias
kp/=scale
else:
kp[:,0] -= self.W_bias
kp[:,1] -= self.H_bias
return {'img': img, 'kp' : kp}
def lookup_cache(self, idx):
if self.cfg.TRAIN.CACHING:
if idx in self.cache_dict:
return self.cache_dict[idx]
sample = self.get_data(idx)
self.cache_dict[idx] = sample
return sample
else:
return self.get_data(idx)
def __getitem__(self, idx):
if self.mode == "train":
while True:
unpair_idx = torch.randint(0, self.len,[1])[0]
if unpair_idx != idx:
break
src = self.lookup_cache(idx)
tgt = self.lookup_cache(unpair_idx)
src_in = src["img"]
target = tgt["img"]
kp_src = src["kp"]
kp_tgt = tgt["kp"]
trans_in = get_limb_transforms(self.limbs, kp_src, kp_tgt)
if self.mode == "val":
unpair_idx = 0
src = self.lookup_cache(unpair_idx)
tgt = self.lookup_cache(idx)
src_in = src["img"]
target = tgt["img"]
kp_src = src["kp"]
kp_tgt = tgt["kp"]
trans_in = get_limb_transforms(self.limbs, kp_src, kp_tgt)
sample = {
'src_in': src_in,
'kp_src': kp_src,
'kp_tgt': kp_tgt,
'trans_in': trans_in,
'target': target
}
return sample
if __name__ == "__main__":
batch_size = 8
dataset = WarpDataset("configs/yaml/oliver.yaml",mode = "train")
print(dataset.__len__())
dataloader = DataLoader(dataset, batch_size=batch_size,
shuffle=False, num_workers=8)
for data in dataloader:
for key in data.keys():
print(key, data[key].shape)
break