-
Notifications
You must be signed in to change notification settings - Fork 0
/
frameworks.py
166 lines (140 loc) · 6.08 KB
/
frameworks.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
"""
functions to quickly train and test localization techniques
"""
from numpy.lib.twodim_base import tri
from helpers import get_visible_waps, make_images, split_frame
from seth.Mapping.Floorplan import Floorplan
from seth.Seth import MAC_RE, Devices, fetch_seth
from stone import (
TripletManager as StoneTripletManager,
embedding_model as stone_embedding_model,
complete_model as stone_complete_model,
pick_negative_1d as stone_pick_negative_1d,
)
from stone.siamese import load_encoder as stone_load_encoder
from paris import (
TripletManager as ParisTripletManager,
embedding_model as paris_embedding_model,
complete_model as paris_complete_model,
pick_negative_1d as paris_pick_negative_1d,
)
from paris.siamese import load_encoder, save_encoder
from paris.attention import (
ParisAttention as ParisAttention,
ParisMultiHeadAttention as ParisMultiHeadAttention,
)
from tensorflow.keras.callbacks import EarlyStopping
from tensorflow_addons.callbacks import TQDMProgressBar
from lt import LTKNN
import numpy as np
from sklearn.neighbors import KNeighborsClassifier as KNN
from data_helper import get_aps_generic as get_aps
from lt import LTKNN
import numpy as np
from sklearn.neighbors import KNeighborsClassifier as KNN
from data_helper import get_aps_generic as get_aps
from tensorflow.keras.models import load_model
import pandas as pd
from typing import List
import tensorflow as tf
from tensorflow_addons.callbacks import TQDMProgressBar
# ######################################
# Stone
# ######################################
def stone_train(train_df,
val_df=None,
target=["label"],
dim_embed=3,
input_shape=(18, 18, 1),
learning_rate=1e-4,
alpha=0.50,
batch_size=32,
steps_per_epoch=100,
p_turn_off=0.80,
contrast_range=None,
brightness_delta=None,
gaussian_noise=0.10,
model_layers=[50, 50, 100],
epochs=100,
callback_loss_patience=20,
fit_verbose=0,
nn=1,
val_bs=100,
encoder_path=None):
if encoder_path is not None:
triplet_encoder, meta = stone_load_encoder(encoder_path)
train_waps = list(meta["TRAIN_WAPS"])
else:
# get train aps
train_waps = get_aps(train_df.columns)
if val_df is None:
train_df, val_df = split_frame(train_df)
# set up train data
train_x = (train_df[train_waps].values + 100) / 100
train_x = make_images(train_x.astype(float), force_shape=input_shape[:2])
# set outputs
train_y = train_df[target].values.reshape((-1)).astype(int)
# setup val data
# set up train data
val_x = (val_df[train_waps].values + 100) / 100
val_x = make_images(val_x.astype(float), force_shape=input_shape[:2])
# set outputs
val_y = val_df[target].values.reshape((-1)).astype(int)
if encoder_path is None:
triplet_encoder = stone_embedding_model(input_shape,
dim_embed,
gaussian_noise=gaussian_noise,
model_layers=model_layers)
# put the encoder into the stone system
siamese = stone_complete_model(triplet_encoder, input_shape, learning_rate, alpha)
# setup data generators and feed the monster!
train_gen = StoneTripletManager(train_x,
train_y,
n_sampler=stone_pick_negative_1d,
steps_per_epoch=steps_per_epoch,
p_turn_off=p_turn_off,
contrast_range=contrast_range,
brightness_delta=brightness_delta,
bs=batch_size)
# validation data generator
val_gen = StoneTripletManager(val_x,
val_y,
n_sampler=stone_pick_negative_1d,
steps_per_epoch=steps_per_epoch,
p_turn_off=p_turn_off,
contrast_range=contrast_range,
brightness_delta=brightness_delta,
bs=val_bs)
# fit the model
history = siamese.fit(train_gen,
validation_data=val_gen,
epochs=epochs,
verbose=fit_verbose,
callbacks=[
EarlyStopping(monitor="val_loss",
patience=callback_loss_patience,
restore_best_weights=True),
TQDMProgressBar(show_epoch_progress=False,
leave_overall_progress=False)
])
print("Loss:", history.history["loss"][-1])
predictor = KNN(nn)
# NOTE: Is it better to train using Augmented Data?
train_encodings = triplet_encoder.predict(train_x)
predictor.fit(train_encodings, train_y.flatten())
return triplet_encoder, predictor, train_waps
def stone_test(test_df, encoder, predictor, train_waps, input_shape=(18, 18, 1)):
# any waps mising can be fixed
test_waps = get_aps(list(test_df.columns))
missing_waps = set(train_waps) - set(test_waps)
test_df[list(missing_waps)] = -100
# tx
tx = np.array(test_df[train_waps].values, dtype=np.float)
tx = (tx + 100) / 100
tx = make_images(tx, force_shape=input_shape[:2])
# test_y = np.array(test_df["label"].values, dtype=np.int).flatten()
# test_y = test_y.reshape((-1, 1))
# predict test using "train_waps"
encoded = encoder.predict(tx)
return predictor.predict(encoded).flatten()
# #####################################