-
Notifications
You must be signed in to change notification settings - Fork 7
/
keras_inference.py
142 lines (112 loc) · 5.48 KB
/
keras_inference.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
from keras.models import load_model
import numpy as np
import pandas as pd
from tqdm import tqdm
import os
from keras import backend as K
from config import *
## params
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
test_dir = 'data/sec_b_fir_sugq_sec_a_sugq_tra_std'
model_file = [
# sec_b > 0.810
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.171-0.8141.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.180-0.8111.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.165-0.8156.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.150-0.8131.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.153-0.8142.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.196-0.8134.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.202-0.8114.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.220-0.8127.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.236-0.8121.hdf5',
# sec_b > 0.809
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.175-0.8097.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.162-0.8096.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.184-0.8093.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.168-0.8093.hdf5',
'model/sec_b_fir_sugq_sec_a_sugq_tra_std/weights.215-0.8098.hdf5',
# # sec_a > 0.830
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.165-0.8316.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.166-0.8303.hdf5',
# # sec_a > 0.829
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.134-0.8291.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.136-0.8298.hdf5',
# # 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.199-0.8298.hdf5',
# # 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.201-0.8297.hdf5',
# # sec_a > 0.828
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.135-0.8286.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.153-0.8282.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.157-0.8282.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.163-0.8286.hdf5',
# # sec_a > 0.827
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.142-0.8275.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.172-0.8270.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.175-0.8273.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.213-0.8270.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.171-0.8273.hdf5',
# 'model/sec_a_fir_sugq_sec_b_sugq_tra_std/weights.184-0.8275.hdf5',
]
## params
def f1(y_true, y_pred):
def recall(y_true, y_pred):
"""Recall metric.
Only computes a batch-wise average of recall.
Computes the recall, a metric for multi-label classification of
how many relevant items are selected.
"""
true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)), axis=0)
possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)), axis=0)
recall = true_positives / (possible_positives + K.epsilon())
return recall
def precision(y_true, y_pred):
"""Precision metric.
Only computes a batch-wise average of precision.
Computes the precision, a metric for multi-label classification of
how many selected items are relevant.
"""
true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)), axis=0)
predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)), axis=0)
precision = true_positives / (predicted_positives + K.epsilon())
return precision
precision = precision(y_true, y_pred)
recall = recall(y_true, y_pred)
return K.sum(2 * ((precision * recall) / (precision + recall + 1e-7))) / 4.0
def most_common_element(array):
(values, counts) = np.unique(array, return_counts=True)
ind = np.argmax(counts)
return values[ind]
def load_test(dir):
test_x = np.load(dir + '/test_x.npy')
test_x = np.expand_dims(test_x, axis=-1)
return test_x
def main():
test_x = load_test(test_dir)
df = pd.read_csv(test_csv).as_matrix()
test_y = []
for mf in tqdm(model_file):
model = load_model(mf, custom_objects={'f1': f1})
y = model.predict(test_x, batch_size=512, verbose=0)
test_y.append(y)
# save logits
# csv_logits = np.hstack((df, y))
# csv_logits = pd.DataFrame(csv_logits)
# csv_logits.to_csv('result/' + mf.split('/')[-1][:-5] + '.csv',
# header=['id', 'star', 'unknown', 'galaxy', 'qso'],
# index=False)
sum_y = np.sum(np.array(test_y), axis=0)
vote_y_idx = np.argmax(sum_y, axis=-1)
vote_y = [label_map_inv[e] for e in vote_y_idx]
# save merged logits
# mean_y = sum_y / len(model_file)
# merge_y = np.hstack((df, mean_y, np.expand_dims(vote_y, axis=-1)))
# pd.DataFrame(merge_y).to_csv('result/pred_lgt.csv', header=['id', 'star', 'unknown', 'galaxy', 'qso', 'pred'], index=False)
# generate csv result
csv_content = np.hstack((df, np.expand_dims(vote_y, axis=-1)))
csv_content = pd.DataFrame(csv_content)
from datetime import datetime
name = datetime.now().strftime('submit/submit_%Y%m%d_%H%M%S') + '.csv'
csv_content.to_csv(name, header=False, index=False)
print('before correction')
print csv_content[1].value_counts()
if __name__ == '__main__':
main()