-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathgenerate_label.py
222 lines (204 loc) · 8.44 KB
/
generate_label.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
import argparse
import glob
import importlib
import os
import shutil
import numpy as np
import ast
from sklearn.cluster import KMeans
from sklearn.mixture import GaussianMixture
import tensorflow as tf
from data import load_data, my_get_subject_files
# from model import TinySleepNet
from minibatching import (iterate_batch_seq_minibatches,
iterate_batch_multiple_seq_minibatches)
from utils import print_n_samples_each_class
from extract_feature import extract_feature
def feature_extract(
config_file,
model_output_dir,
data_output_dir,
model_based=False
):
spec = importlib.util.spec_from_file_location("*", config_file)
config = importlib.util.module_from_spec(spec)
spec.loader.exec_module(config)
config = config.predict
# Create output directory, model is not writing anything to this dir in this function
if not os.path.exists(model_output_dir):
os.makedirs(model_output_dir)
subject_files = glob.glob(os.path.join(config["feature_extract_file_dir"], "*.npz"))
# Add dummy class weights
config["class_weights"] = np.ones(config["n_classes"], dtype=np.float32)
# model = TinySleepNet(
# config=config,
# output_dir=model_output_dir,
# use_rnn=True,
# testing=True,
# use_best=True,
# feature_extract=True
# )
test_files = my_get_subject_files(files=subject_files)
test_x, test_y, fs, _ = load_data(test_files, data_from_cluster=False)
# Print test set
print_n_samples_each_class(np.hstack(test_y))
logits = extract_feature(test_x, fs)
labels = []
for i in range(len(test_y)):
for j in test_y[i]:
labels.append(j)
ground_truth = labels
# logits, labels, ground_truth = pred_logits(config, test_x, test_y, model)
generated_labels = []
class_num = 5
# clustering
if model_based:
cluster_train_subject_files = glob.glob(os.path.join(config["feature_extract_train_file_dir"], "*.npz"))
train_files = my_get_subject_files(files=cluster_train_subject_files)
cluster_train_X, cluster_train_y, fs1, _ = load_data(train_files, data_from_cluster=False)
cluster_train_logits=extract_feature(cluster_train_X, fs1)
# cluster_train_logits, _, _ = pred_logits(config, cluster_train_X, cluster_train_y, model)
cluster_label = clustering(cluster_train_logits, logits, class_num, True)
else:
cluster_label = clustering(None, logits, class_num, False)
label_map = np.zeros((class_num, class_num))
label_map_true = np.zeros((class_num, class_num))
for i in range(cluster_label.shape[0]):
label_map[cluster_label[i], labels[i]] += 1
print(label_map)
print()
for i in range(cluster_label.shape[0]):
label_map_true[cluster_label[i], ground_truth[i]] += 1
print(label_map_true)
# correct labels, make data in the same cluster have the same label
corrected_label = -1 * np.ones((label_map.shape[0],))
ratio = np.zeros((class_num, class_num))
for i in range(label_map.shape[0]):
cluster_sum = np.sum(label_map[i])
ratio[i] = label_map[i]/cluster_sum
print(ratio)
for i in range(class_num):
# corrected_label[np.where(ratio[:, i] == np.amax(ratio[:, i]))[0][0]] = i
corrected_label[i] = np.where(ratio[i] == np.amax(ratio[i]))[0][0]
print(corrected_label)
for i in range(cluster_label.shape[0]):
generated_labels.append(int((corrected_label[cluster_label[i]])))
assert len(labels) == len(generated_labels)
# split data into n_folds files
n_folds = 6
len_per_file = len(labels)//n_folds
flatten_x = []
for i in range(len(test_x)):
for j in range(test_x[i].shape[0]):
flatten_x.append(test_x[i][j])
new_x = []
new_cluster_label = []
new_true_label = []
start = 0
correct_num = 0
total_num = 0
for i in range(n_folds):
x = []
y = []
y_cluster = []
if start+len_per_file <= len(flatten_x):
end = start+len_per_file
else:
end = len(flatten_x)
x.extend(flatten_x[start:end])
y.extend(ground_truth[start:end])
y_cluster.extend((generated_labels[start:end]))
start += len_per_file
new_x.append(np.array(x))
new_true_label.append(np.array(y))
new_cluster_label.append((np.array(y_cluster)))
correct_num +=np.sum(np.array(y) == np.array(y_cluster))
total_num +=len(x)
print(correct_num)
print(total_num)
print(correct_num/total_num)
# Output dir
# if not os.path.exists(data_output_dir):
# os.makedirs(data_output_dir)
# else:
# shutil.rmtree(data_output_dir)
# os.makedirs(data_output_dir)
# for i in range(len(new_true_label)):
# # Save
# filename = f"{str(i).zfill(2)}.npz"
# save_dict = {
# "x": new_x[i].astype(np.float32),
# "y": new_true_label[i].astype(np.int32),
# "y_cluster": new_cluster_label[i].astype(np.int32),
# "fs": 100
# }
# np.savez(os.path.join(data_output_dir, filename), **save_dict)
return
def pred_logits(config, X, y, model):
"""Given signal, generate labels using tinysleepnet"""
logits = []
labels = []
ground_truth = []
if config["model"] == "model-origin":
for night_idx, night_data in enumerate(zip(X, y)):
# Create minibatches for testing
night_x, night_y = night_data
test_minibatch_fn = iterate_batch_seq_minibatches(
night_x,
night_y,
batch_size=config["batch_size"],
seq_length=config["seq_length"],
)
# Evaluate
test_outs = model.evaluate(test_minibatch_fn)
logits.extend(test_outs["test/logits"])
labels.extend(test_outs["test/preds"])
ground_truth.extend(test_outs["test/trues"])
else:
for night_idx, night_data in enumerate(zip(X, y)):
# Create minibatches for testing
night_x, night_y = night_data
test_minibatch_fn = iterate_batch_multiple_seq_minibatches(
[night_x],
[night_y],
batch_size=config["batch_size"],
seq_length=config["seq_length"],
shuffle_idx=None,
augment_seq=False,
)
if (config.get('augment_signal') is not None) and config['augment_signal']:
# Evaluate
test_outs = model.evaluate_aug(test_minibatch_fn)
else:
# Evaluate
test_outs = model.evaluate(test_minibatch_fn)
logits.extend(test_outs["test/logits"])
labels.extend(test_outs["test/preds"])
ground_truth.extend(test_outs["test/trues"])
tf.reset_default_graph()
return logits, labels, ground_truth
def clustering(train_X, pred_X, class_num, model_based):
if model_based:
train_X = np.squeeze(train_X)
gm = GaussianMixture(n_components=class_num, random_state=0).fit(train_X)
cluster_label = gm.predict(np.squeeze(pred_X))
else:
cluster_label = KMeans(n_clusters=class_num, random_state=1).fit(pred_X).labels_
return cluster_label
if __name__ == "__main__":
# parser = argparse.ArgumentParser()
# parser.add_argument("--config_file", type=str, required=True)
# parser.add_argument("--model_based", type=ast.literal_eval, required=True)
# parser.add_argument("--model_output_dir", type=str, default="./out_sleepedf/feature_extract")
# parser.add_argument("--data_output_dir", type=str, default="./my_data/labels_from_cluster_data16")
# args = parser.parse_args()
# # --config_file config/cluster.py
# # --model_output_dir out_sleepedf/feature_extract
# feature_extract(config_file=args.config_file,
# model_output_dir=args.model_output_dir,
# data_output_dir=args.data_output_dir,
# model_based=args.model_based)
feature_extract(config_file="config/cluster.py",
model_output_dir="./out_sleepedf/feature_extract",
data_output_dir="./my_data/labels_from_cluster_data16",
model_based=True)