-
Notifications
You must be signed in to change notification settings - Fork 16
/
dataset.py
309 lines (243 loc) · 10.8 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
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
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
import torch
import numpy as np
import pickle
import random
SKETCH_R = 1
RADIUS_R = 1
EXTRUDE_R = 1.0
SCALE_R = 1.4
OFFSET_R = 0.9
PIX_PAD = 4
CMD_PAD = 3
COORD_PAD = 4
EXT_PAD = 1
EXTRA_PAD = 1
R_PAD = 2
AUG_RANGE = 5
MAX_EXT = 5
class SketchData(torch.utils.data.Dataset):
""" sketch dataset """
def __init__(self, data_path, invalid_uid, MAX_LEN):
self.maxlen = MAX_LEN
self.maxlen_pix = 0
self.maxlen_cmd = 0
self.maxlen_ext = 0
self.maxlen_se = 0
with open(invalid_uid, 'rb') as f:
invalid_uids = pickle.load(f)
invaliduid = {}
for invalid in invalid_uids:
invaliduid[invalid] = True
######################
## Load sketch data ##
######################
with open(data_path, 'rb') as f:
data = pickle.load(f)
self.data = {}
for index in range(len(data)):
vec_data = data[index]
pix_len = vec_data['len_pix']
cmd_len = vec_data['len_cmd']
ext_len = vec_data['len_ext']
num_se = vec_data['num_se']
total_len = pix_len + EXTRA_PAD
uid = vec_data['name']
if total_len <= self.maxlen and uid not in invaliduid:
self.data[uid] = vec_data
if pix_len+EXTRA_PAD > self.maxlen_pix:
self.maxlen_pix = pix_len+EXTRA_PAD
if cmd_len+EXTRA_PAD > self.maxlen_cmd:
self.maxlen_cmd = cmd_len+EXTRA_PAD
if ext_len+EXTRA_PAD > self.maxlen_ext:
self.maxlen_ext = ext_len+EXTRA_PAD
if num_se > self.maxlen_se:
self.maxlen_se = num_se
self.uids = sorted(list(set(self.data.keys()).intersection(set(self.data.keys()))))
# print(f'Sketch Post-Filter: {len(self.uids)}, Keep Ratio: {100*len(self.uids)/len(data):.2f}%')
# print(f'Max Pix {self.maxlen_pix}, Max CMD {self.maxlen_cmd}, Max SE {self.maxlen_se}')
def __len__(self):
return len(self.uids)
def prepare_batch_sketch(self, pixel_v, xy_v):
keys = np.ones(len(pixel_v))
padding = np.zeros(self.maxlen_pix-len(pixel_v)).astype(int)
pixel_v_flat = np.concatenate([pixel_v, padding], axis=0)
pixel_v_mask = 1-np.concatenate([keys, padding]) == 1
padding = np.zeros((self.maxlen_pix-len(xy_v), 2)).astype(int)
xy_v_flat = np.concatenate([xy_v, padding], axis=0)
return pixel_v_flat, xy_v_flat, pixel_v_mask
def prepare_batch_cmd(self, command):
keys = np.ones(len(command))
padding = np.zeros(self.maxlen_cmd-len(command)).astype(int)
command_pad = np.concatenate([command, padding])
mask = 1-np.concatenate([keys, padding]) == 1
return command_pad, mask
def __getitem__(self, index):
uid = self.uids[index]
vec_data = self.data[uid]
pix_tokens = vec_data['se_pix']
xy_tokens = vec_data['se_xy']
cmd_tokens = vec_data['se_cmd']
pixs = np.hstack(pix_tokens)+EXTRA_PAD
pixs = np.concatenate((pixs, np.zeros(1).astype(int)))
xys = np.vstack(xy_tokens)+EXTRA_PAD
xys = np.concatenate((xys, np.zeros((1,2)).astype(int)))
cmds = np.hstack(cmd_tokens)+EXTRA_PAD
cmds = np.concatenate((cmds, np.zeros(1).astype(int)))
pix_seq, xy_seq, mask = self.prepare_batch_sketch(pixs, xys)
cmd_seq, cmd_mask = self.prepare_batch_cmd(cmds)
# Quantization augmentation
aug_xys = []
for xy in xys:
if xy[0] <= COORD_PAD and xy[1] <= COORD_PAD:
aug_xys.append(xy - COORD_PAD - EXTRA_PAD)
else:
new_xy = xy - COORD_PAD - EXTRA_PAD
new_xy[0] = new_xy[0] + random.randint(-AUG_RANGE, +AUG_RANGE)
new_xy[1] = new_xy[1] + random.randint(-AUG_RANGE, +AUG_RANGE)
new_xy = np.clip(new_xy, a_min=0, a_max=2**6-1)
aug_xys.append(new_xy)
_xys_ = np.vstack(aug_xys) + EXTRA_PAD + COORD_PAD
# # Augment the pix value according to XY
aug_pix = []
for xy in aug_xys:
if xy[0] >= 0 and xy[1] >= 0:
aug_pix.append(xy[1]*(2**6)+xy[0])
else:
aug_pix.append(xy[0])
_pixs_ = np.hstack(aug_pix) + EXTRA_PAD + PIX_PAD
pix_seq_aug, xy_seq_aug, mask_aug = self.prepare_batch_sketch(_pixs_, _xys_)
return cmd_seq, cmd_mask, pix_seq, xy_seq, mask, \
pix_seq_aug, xy_seq_aug, mask_aug
class CodeDataset(torch.utils.data.Dataset):
""" Code dataset """
def __init__(self, datapath, maxlen):
with open(datapath, 'rb') as f:
self.data = pickle.load(f)
self.maxlen = maxlen
print(len(self.data))
return
def __len__(self):
return len(self.data)
def __getitem__(self, index):
code = self.data[index]
return code
class SketchExtData(torch.utils.data.Dataset):
""" sketch dataset """
def __init__(self, data, invalid_uid, MAX_LEN):
self.maxlen = MAX_LEN
self.maxlen_pix = 0
self.maxlen_cmd = 0
self.maxlen_ext = 0
# Convert list to dictionary, signicantly faster for key indexing
with open(invalid_uid, 'rb') as f:
invalid_uids = pickle.load(f)
invaliduid = {}
for invalid in invalid_uids:
invaliduid[invalid] = True
######################
## Load sketch data ##
######################
with open(data, 'rb') as f:
data = pickle.load(f)
self.data = {}
for index in range(len(data)):
vec_data = data[index]
pix_len = vec_data['len_pix']
cmd_len = vec_data['len_cmd']
ext_len = vec_data['len_ext']
total_len = pix_len + EXTRA_PAD
uid = vec_data['name']
ext_len = vec_data['len_ext']
if total_len <= self.maxlen and vec_data['num_se']<=MAX_EXT and uid not in invaliduid:
self.data[uid] = vec_data
ext_len = vec_data['len_ext']
if pix_len+EXTRA_PAD > self.maxlen_pix:
self.maxlen_pix = pix_len+EXTRA_PAD
if cmd_len+EXTRA_PAD > self.maxlen_cmd:
self.maxlen_cmd = cmd_len+EXTRA_PAD
if ext_len+EXTRA_PAD > self.maxlen_ext:
self.maxlen_ext = ext_len+EXTRA_PAD
self.uids = sorted(list(set(self.data.keys())))
# print(f'Sketch Post-Filter: {len(self.uids)}, Keep Ratio: {100*len(self.uids)/len(data):.2f}%')
# print(f'Max Pix {self.maxlen_pix}, Max CMD {self.maxlen_cmd}')
def __len__(self):
return len(self.uids)
def prepare_batch_sketch(self, pixel_v, xy_v):
keys = np.ones(len(pixel_v))
padding = np.zeros(self.maxlen_pix-len(pixel_v)).astype(int)
pixel_v_flat = np.concatenate([pixel_v, padding], axis=0)
pixel_v_mask = 1-np.concatenate([keys, padding]) == 1
padding = np.zeros((self.maxlen_pix-len(xy_v), 2)).astype(int)
xy_v_flat = np.concatenate([xy_v, padding], axis=0)
return pixel_v_flat, xy_v_flat, pixel_v_mask
def prepare_batch_cmd(self, command):
keys = np.ones(len(command))
padding = np.zeros(self.maxlen_cmd-len(command)).astype(int)
command_pad = np.concatenate([command, padding])
mask = 1-np.concatenate([keys, padding]) == 1
return command_pad, mask
def prepare_batch_extrude(self, ext, flags):
keys = np.ones(len(ext))
padding = np.zeros(self.maxlen_ext-len(ext)).astype(int)
flag_pad = np.concatenate([flags, padding], axis=0)
ext_flat = np.concatenate([ext, padding], axis=0)
ext_mask = 1-np.concatenate([keys, padding]) == 1
return ext_flat, flag_pad, ext_mask
def __getitem__(self, index):
uid = self.uids[index]
vec_data = self.data[uid]
pix_tokens = vec_data['se_pix']
xy_tokens = vec_data['se_xy']
cmd_tokens = vec_data['se_cmd']
ext_tokens = vec_data['se_ext']
exts = np.hstack(ext_tokens) + EXTRA_PAD
exts = np.concatenate((exts, np.zeros(1).astype(int)))
ext_flags = np.hstack([1,1,2,2,2,3,3,3,3,3,3,3,3,3,4,5,6,6,7] * len(ext_tokens))
ext_flags = np.concatenate((ext_flags, np.zeros(1).astype(int)))
ext_seq, flag_seq, ext_mask = self.prepare_batch_extrude(exts, ext_flags)
pixs = np.hstack(pix_tokens)+EXTRA_PAD
pixs = np.concatenate((pixs, np.zeros(1).astype(int)))
xys = np.vstack(xy_tokens)+EXTRA_PAD
xys = np.concatenate((xys, np.zeros((1,2)).astype(int)))
cmds = np.hstack(cmd_tokens)+EXTRA_PAD
cmds = np.concatenate((cmds, np.zeros(1).astype(int)))
pix_seq, xy_seq, sketch_mask = self.prepare_batch_sketch(pixs, xys)
cmd_seq, cmd_mask = self.prepare_batch_cmd(cmds)
return cmd_seq, cmd_mask, pix_seq, xy_seq, sketch_mask, flag_seq, ext_seq, ext_mask
class ExtData(torch.utils.data.Dataset):
""" extrude dataset """
def __init__(self, data_path, MAX_LEN):
with open(data_path, 'rb') as f:
data = pickle.load(f)
self.maxlen = MAX_LEN
self.maxlen_ext = 0
# Filter out too long results
self.data = []
for index in range(len(data)):
vec_data = data[index]
uid = vec_data['name']
if vec_data['num_se'] <= self.maxlen:
self.data.append(vec_data)
ext_len = vec_data['len_ext']
if ext_len+EXTRA_PAD > self.maxlen_ext:
self.maxlen_ext = ext_len+EXTRA_PAD
# print(f'Sketch Post-Filter: {len(self.data)}, Keep Ratio: {100*len(self.data)/len(data):.2f}%')
# print(f'Max Ext {self.maxlen_ext}')
def __len__(self):
return len(self.data)
def prepare_batch_extrude(self, ext, flags):
keys = np.ones(len(ext))
padding = np.zeros(self.maxlen_ext-len(ext)).astype(int)
flag_pad = np.concatenate([flags, padding], axis=0)
ext_flat = np.concatenate([ext, padding], axis=0)
ext_mask = 1-np.concatenate([keys, padding]) == 1
return ext_flat, flag_pad, ext_mask
def __getitem__(self, index):
vec_data = self.data[index]
ext_tokens = vec_data['se_ext']
exts = np.hstack(ext_tokens) + EXT_PAD
exts = np.concatenate((exts, np.zeros(1).astype(int)))
ext_flags = np.hstack([1,1,2,2,2,3,3,3,3,3,3,3,3,3,4,5,6,6,7] * len(ext_tokens))
ext_flags = np.concatenate((ext_flags, np.zeros(1).astype(int)))
ext_seq, flag_seq, ext_mask = self.prepare_batch_extrude(exts, ext_flags)
return ext_seq, flag_seq, ext_mask