-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdataset_nyu_arm_com.py
181 lines (140 loc) · 7.01 KB
/
dataset_nyu_arm_com.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
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
import torch.utils.data as data
import os
import os.path
import torch
import numpy as np
import scipy.io as sio
import pdb
from tqdm import tqdm
SAMPLE_NUM = 1024
JOINT_NUM = 14
class HandPointDatasetArm(data.Dataset):
def __init__(self, root_path, opt, sample=2048, output_num=1024, train=True, shuffle=False):
self.root_path = root_path
self.train = train
self.SAMPLE_NUM = sample
self.OUTPUT_NUM = output_num
self.INPUT_FEATURE_NUM = opt.INPUT_FEATURE_NUM
self.JOINT_NUM = opt.JOINT_NUM
self.restrictedJointsEval = [0, 3, 6, 9, 12, 15, 18, 21, 24, 25, 27, 30, 31, 32]
self.record_file, self.record_data = self.__fileToNumpy(os.path.join(root_path, 'record.txt'))
self.total_frame_num = len(self.record_file)
print(self.total_frame_num)
self.point_clouds = np.empty(shape=[self.total_frame_num, self.SAMPLE_NUM, self.INPUT_FEATURE_NUM],
dtype=np.float32)
self.volume_length = np.empty(shape=[self.total_frame_num, 1], dtype=np.float32)
self.gt_xyz = np.empty(shape=[self.total_frame_num, 36*3], dtype=np.float32)
self.offset = np.empty(shape=[self.total_frame_num, 3], dtype=np.float32)
self.rotate = np.empty(shape=[self.total_frame_num, 3, 3], dtype=np.float32)
self.start_index = 0
self.end_index = 0
print("Loading Dataset..........")
for i in tqdm(range(self.total_frame_num)):
cur_data_dir = os.path.join(self.root_path, self.record_file[i] + '_Point_Cloud_FPS.mat')
# print("Training: " + cur_data_dir)
self.__loaddata(cur_data_dir)
self.gt_xyz = self.record_data[:, 1:109].astype(np.float32)
self.volume_length = self.record_data[:, 0].astype(np.float32)
self.offset = self.record_data[:, 109:112].astype(np.float32)
self.rotate = self.record_data[:, 112:].astype(np.float32).reshape(-1,3,3)
if shuffle:
idx_shuffle = np.random.permutation(len(self.point_clouds))
self.point_clouds = self.point_clouds[idx_shuffle]
self.volume_length = self.volume_length[idx_shuffle]
self.gt_xyz = self.gt_xyz[idx_shuffle]
self.offset = self.offset[idx_shuffle]
self.rotate = self.rotate[idx_shuffle]
self.point_clouds = torch.from_numpy(self.point_clouds)
self.volume_length = torch.from_numpy(self.volume_length).view(self.total_frame_num, 1)
self.gt_xyz = torch.from_numpy(self.gt_xyz)
self.offset = torch.from_numpy(self.offset)
self.rotate = torch.from_numpy(self.rotate)
self.total_frame_num = self.point_clouds.size(0)
def __getitem__(self, index):
# sample_idx = np.random.choice(self.SAMPLE_NUM, 1024)
# return self.point_clouds[index, sample_idx, :], self.volume_length[index], self.gt_xyz.view(-1,36,3), self.offset[index],
if self.OUTPUT_NUM != self.SAMPLE_NUM:
pts = self.__random_cut_arm(self.point_clouds[index, :, :], self.rotate[index], self.gt_xyz[index, :].view(-1, 3), self.OUTPUT_NUM)
# pts = self.__random_cut_arm(self.point_clouds[index, :, :], self.rotate[index], self.gt_xyz[index, :].view(-1, 3), self.OUTPUT_NUM)
gt_xyz = self.gt_xyz[index, :].view(-1, 3)
offset = self.offset[index]
else:
pts = self.point_clouds[index, :, :]
offset = self.offset[index]
gt_xyz = self.gt_xyz[index, :].view(-1, 3)
return pts, self.volume_length[index], gt_xyz[self.restrictedJointsEval,:].view(14*3), offset
def __len__(self):
return self.point_clouds.size(0)
def __loaddata(self, data_dir):
point_cloud = sio.loadmat(data_dir)
self.start_index = self.end_index + 1
self.end_index = self.end_index + 1
self.point_clouds[(self.start_index - 1):self.end_index, :, :] = point_cloud['Point_Cloud_FPS'].astype(np.float32)[:,:3]
def __random_cut_arm(self, points, rot, output_num):
points = torch.matmul(points, rot)
max_ = torch.max(points, 0)[0]
min_ = torch.min(points, 0)[0]
box = max_ - min_
if box[0] > box[1]:
mask = points[:,0] > (max_[0] - 0.8 - np.random.rand() * (max_[0] - min_[0] - 0.8))
sampled_points = points[mask]
idx = np.random.choice(len(sampled_points), output_num)
points = sampled_points[idx]
elif box[1] > box[0]:
mask1 = points[:,1] > (-0.4 - np.random.rand() * (-0.4 - min_[1]))
mask2 = points[:,1] < (0.4 + np.random.rand() * (max_[1] - 0.4))
mask = mask1 & mask2
sampled_points = points[mask]
idx = np.random.choice(len(sampled_points), output_num)
points = sampled_points[idx]
else:
idx = np.random.choice(len(points), output_num)
points = points[idx]
points = torch.matmul(points, rot.T)
return points
def __random_cut_arm(self, points, rot, joints, output_num):
points = torch.matmul(points, rot)
joints = torch.matmul(joints.view(-1, 3), rot).view(-1)
anchors = joints[[30, 31]]
max_anchor = torch.min(anchors, 0)[0]
max_ = torch.max(points, 0)[0]
min_ = torch.min(points, 0)[0]
box = max_ - min_
if box[0] > box[1]:
mask = points[:,0] > (max_anchor - np.random.rand() * (max_anchor - min_[0]))
sampled_points = points[mask]
if len(sampled_points) > output_num:
idx = np.random.choice(len(sampled_points), output_num)
points = sampled_points[idx]
else:
idx = np.random.choice(len(points), output_num)
points = points[idx]
elif box[1] > box[0]:
mask1 = points[:,1] > (-0.3 - np.random.rand() * (-0.3 - min_[1]))
mask2 = points[:,1] < (0.3 + np.random.rand() * (max_[1] - 0.3))
mask = mask1 & mask2
sampled_points = points[mask]
idx = np.random.choice(len(sampled_points), output_num)
points = sampled_points[idx]
else:
idx = np.random.choice(len(points), output_num)
points = points[idx]
points = torch.matmul(points, rot.T)
return points
def __get_frmae_num(self, data_dir):
volume_length = sio.loadmat(os.path.join(data_dir, "Volume_length.mat"))
return len(volume_length['Volume_length'])
def __fileToNumpy(self, filename):
file = open(filename)
file_lines = file.readlines()
numberOfLines = len(file_lines)
dataArray = np.zeros((numberOfLines, 121))
labels = []
index = 0
for line in file_lines:
line = line.strip() # 参数为空时，默认删除开头、结尾处空白符（包括'\n', '\r', '\t', ' ')
formLine = line.split(' ')
dataArray[index,:] = formLine[1:122]
labels.append((formLine[0]))
index += 1
return labels, dataArray