/
create_experiment.py
258 lines (199 loc) · 8.58 KB
/
create_experiment.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
import argparse
import json
import re
import h5py
import numpy as np
import spacy
from scipy import spatial
import os #add
KITCHEN_ID = 0
LIVINGROOM_ID = 200
BEDROOM_ID = 300
BATHROOM_ID = 400
names = []
# SCENES_TRAINING = [KITCHEN_ID, BEDROOM_ID] # origin
SCENES_TRAINING = [KITCHEN_ID, LIVINGROOM_ID, BEDROOM_ID, BATHROOM_ID] # add
# SCENES_EVAL = [LIVINGROOM_ID, BATHROOM_ID] # origin
SCENES_EVAL = [KITCHEN_ID, LIVINGROOM_ID, BEDROOM_ID, BATHROOM_ID] #add
TRAIN_SPLIT = (1, 21)
TEST_SPLIT = (22, 27)
KITCHEN_OBJECT_CLASS_LIST_TRAIN = [
"Toaster",
"Microwave",
"Fridge",
"CoffeeMachine",
"GarbageCan",
"Bowl",
]
KITCHEN_OBJECT_CLASS_LIST_EVAL = [
"Mug",
"Pot",
"Cup"
]
LIVING_ROOM_OBJECT_CLASS_LIST_TRAIN = [
"Pillow",
"Laptop",
"Television",
"GarbageCan",
"Bowl",
]
LIVING_ROOM_OBJECT_CLASS_LIST_EVAL = [
"Sofa",
"Box",
"TableTop"
]
BEDROOM_OBJECT_CLASS_LIST_TRAIN = ["HousePlant", "Lamp", "Book", "AlarmClock"]
BEDROOM_OBJECT_CLASS_LIST_EVAL = ["Mirror", "CD", "CellPhone"]
BATHROOM_OBJECT_CLASS_LIST_TRAIN = [
"Sink", "ToiletPaper", "SoapBottle", "LightSwitch"]
BATHROOM_OBJECT_CLASS_LIST_EVAL = [
"Toilet", "Towel"]
scene_id_name = ["Kitchen", "LivingRoom", "Bedroom", "Bathroom"]
def extract_word_emb_vector(nlp, word_name):
# Usee scapy to extract word embedding vector
word_vec = nlp(word_name.lower())
# If words don't exist in dataset
# cut them using uppercase letter (SoapBottle -> Soap Bottle)
if word_vec.vector_norm == 0:
word = re.sub(r"(?<=\w)([A-Z])", r" \1", word_name)
word_vec = nlp(word.lower())
# If no embedding found try to cut word to find embedding (SoapBottle -> [Soap, Bottle])
if word_vec.vector_norm == 0:
word_split = re.findall('[A-Z][^A-Z]*', word)
for word in word_split:
word_vec = nlp(word.lower())
if word_vec.has_vector:
break
if word_vec.vector_norm == 0:
print('ERROR: %s not found' % word_name)
return None
norm_word_vec = word_vec.vector / word_vec.vector_norm # Normalize vector size
return norm_word_vec, word_vec.text
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description='Create param.json experiment file')
parser.add_argument('--env', default='',
help='Create a directory and .env') #add Default, Output to EXPERIMENT, Escape override
parser.add_argument('--train_range', nargs=2, default=TRAIN_SPLIT,
help='train scene range Ex : 1 11')
parser.add_argument('--eval_range', nargs=2, default=TEST_SPLIT,
help='train scene range Ex : 22 27')
parser.add_argument('--tstep', default=25000000,
help='total_step Ex : 25000000') #add
parser.add_argument('--period', default=1000000,
help='saving_period Ex : 1000000') #add
parser.add_argument('--max_t', default=5,
help='max_t Ex : 5') #add
parser.add_argument('--actions', default=9,
help='action_size Ex : 9') #add
parser.add_argument('--ngpu', default=4,
help='NGPU(Number of GPUs used) Ex : 4') #add
parser.add_argument('--key', default="word2vec",
help='Key Ex : "word2vec"') #add
parser.add_argument('--memory', default=32,
help='memory size can be changed Ex : 32') #add
parser.add_argument('--bbox_method', default='bbox',
help='bbox_method can be changed : bbox or yolo') #add
parser.add_argument('--thread', default=8,
help='num_thread Ex : 8') #add
parser.add_argument('--gamma', default=0.7,
help='gamma Ex : 0.7') #add
parser.add_argument('--seed', default=1993,
help='seed Ex : 1993') #add
parser.add_argument('--reward', type=str, default="soft_goal",
help='Method to use Ex : soft_goal')
parser.add_argument('--masks', default=16,
help='mask_size Ex : 16') #add
parser.add_argument('--nepi', default=250,
help='num_episode Ex : 250') #add
parser.add_argument('--method', type=str, default="grid_memory",
help='Method to use Ex : grid_memory')
parser.add_argument('--eval_objects', action="store_true")
args_add = parser.parse_args() #add
args = vars(parser.parse_args())
str_range = list(args["train_range"])
for i, s in enumerate(str_range):
str_range[i] = int(s)
args["train_range"] = str_range
str_range = list(args["eval_range"])
for i, s in enumerate(str_range):
str_range[i] = int(s)
args["eval_range"] = str_range
data = {}
scene_tasks = { KITCHEN_ID: KITCHEN_OBJECT_CLASS_LIST_TRAIN,
LIVINGROOM_ID: LIVING_ROOM_OBJECT_CLASS_LIST_TRAIN,
BEDROOM_ID: BEDROOM_OBJECT_CLASS_LIST_TRAIN,
BATHROOM_ID: BATHROOM_OBJECT_CLASS_LIST_TRAIN}
training = {}
set_obj = None
for idx_scene, scene in enumerate(SCENES_TRAINING):
for t in range(*args['train_range']):
name = "FloorPlan" + str(scene + t)
f = h5py.File("data/"+name+".h5", 'r')
# Use h5py object available
obj_available = json.loads(f.attrs["task_present"])
obj_available = np.array(list(set.intersection(
set(obj_available), set(scene_tasks[scene]))))
obj_available = np.array(obj_available)
obj_available_mask = [False for i in obj_available]
obj_available_mask = np.array(obj_available_mask)
object_visibility_tmp = [json.loads(j) for j in
f['object_visibility']]
object_visibility = set()
for obj_visible in object_visibility_tmp:
for objectId in obj_visible:
obj = objectId.split('|')
object_visibility.add(obj[0])
object_visibility = list(object_visibility)
for obj_visible in object_visibility:
for obj_idx, curr_obj in enumerate(obj_available):
if obj_visible == curr_obj:
obj_available_mask[obj_idx] = True
break
training[name] = [{"object": obj}
for obj in obj_available[obj_available_mask == True]]
if args['eval_objects']:
scene_tasks = { KITCHEN_ID: KITCHEN_OBJECT_CLASS_LIST_EVAL,
LIVINGROOM_ID: LIVING_ROOM_OBJECT_CLASS_LIST_EVAL,
BEDROOM_ID: BEDROOM_OBJECT_CLASS_LIST_EVAL,
BATHROOM_ID: BATHROOM_OBJECT_CLASS_LIST_EVAL}
evaluation = {}
evaluation_set = dict()
for idx_scene, scene in enumerate(SCENES_EVAL):
evaluation_set[scene] = list()
for t in range(*args['eval_range']):
name = "FloorPlan" + str(scene + t)
evaluation[name] = [
{"object": obj} for obj in scene_tasks[scene]]
data["task_list"] = {}
data["task_list"]["train"] = training
data["task_list"]["eval"] = evaluation
data["total_step"] = int(args_add.tstep) #origin 25000000
data["h5_file_path"] = "./data/{scene}.h5"
data["saving_period"] = int(args_add.period) #origin 1000000
data["max_t"] = int(args_add.max_t) #add origin 5
data["action_size"] = int(args_add.actions) #add origin 9
data["SSL"] = False #add
data["Posi"] = False #add
data["Key"] = str(args_add.key) #add
data["NGPU"] = int(args_add.ngpu) #add origin 4
data["memory"] = int(args_add.memory) #add
data["bbox_method"] = str(args_add.bbox_method) #add
data["restore"] = False #add
train_param = {}
train_param["cuda"] = True
train_param["num_thread"] = int(args_add.thread) #origin 8
train_param["gamma"] = float(args_add.gamma) #origin 0.7
train_param["seed"] = int(args_add.seed) #origin 1993
train_param["reward"] = args["reward"]
train_param["mask_size"] = int(args_add.masks) #origin 16
data["train_param"] = train_param
data["eval_param"] = {}
data["eval_param"]["num_episode"] = int(args_add.nepi) #origin 250
data["method"] = args["method"]
os.makedirs("EXPERIMENT/" + args_add.env, exist_ok=True) #add
with open('.env', mode='w', encoding='utf-8') as f: #add
f.write(str(args_add.env)) #add
os.chdir("EXPERIMENT/" + args_add.env) #add
with open('param.json', 'w') as outfile:
outfile.write(json.dumps(data, indent=4))