-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrefinementClassification.py
163 lines (133 loc) · 5.69 KB
/
refinementClassification.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
import time
import joblib
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
from joblib import dump
from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor
from sklearn.metrics import (accuracy_score, classification_report,
confusion_matrix)
from RefinedRandomForest import RefinedRandomForest
plt.style.use('ggplot')
def ind2onehot(a):
b = np.zeros((a.size, a.max()+1))
b[np.arange(a.size),a] = 1
return b
def mae(y_true : np.ndarray, y_pred: np.ndarray) -> float:
if y_true.shape[0] == 0:
return 0
return np.abs(y_true - y_pred).mean()
def aar(y_true : np.ndarray, y_pred: np.ndarray) -> float:
true_age_groups = np.clip(y_pred // 10, 0, 7)
mae_score = mae(y_true, y_pred)
# MAE per age group
sigmas = []
maes = []
for i in range(8):
idx = true_age_groups == i
mae_age_group = mae(y_true[idx], y_pred[idx])
maes.append(mae_age_group)
sigmas.append((mae_age_group - mae_score) ** 2)
sigma = np.sqrt(np.array(sigmas).mean())
aar_score = max(0, 7 - mae_score) + max(0, 3 - sigma)
return aar_score, mae_score, sigma, sigmas, maes
if __name__ == '__main__':
df = pd.read_csv('../../training_caip_contest.csv',
header=None)
x = np.load('../../../../mvl5/Images2/DeepFake/GTA/features_pred_resnext_aar.npy')
# x = np.load('data/features_pred_resnext_mse.npy')
y = np.array(df.iloc[:, 1])
train_index = np.loadtxt('../train_index.txt', delimiter=',').astype('int')
test_index = np.loadtxt('../test_index.txt', delimiter=',').astype('int')
x_train = x[train_index]
y_train = y[train_index]
y_train_group = np.clip(y_train // 10, 0, 7)
x_test = x[test_index]
y_test = y[test_index]
y_test_group = np.clip(y_test // 10, 0, 7)
n_estimators = 100
min_samples_leaf = 5
max_features = 128
random_state = 1
rfc = joblib.load('data/tlrf_resnext_rfc_100_5_128_1.joblib')
classify_accuracy = []
weighted_f1 = []
final_aar = []
final_MAE = []
leaves = []
y_train_prob = rfc.predict_proba(x_train)
x_train_cat = np.concatenate([x_train, y_train_prob], axis=1)
outc = rfc.predict_proba(x_test)
y_pred_group = outc.argmax(axis=1)
#print(f"Classifier accuracy: {accuracy_score(y_test_group, y_pred_group):.3f}")
#print(f"Report: \n {classification_report(y_test_group, y_pred_group)}")
classify_accuracy.append(accuracy_score(y_test_group, y_pred_group))
r = classification_report(y_test_group, y_pred_group).split()
weighted_f1.append(r[57])
rfr = joblib.load('data/tlrf_resnext_rfr_100_5_128_1.joblib')
x_test_cat = np.concatenate([x_test, outc], axis=1)
out = np.clip(rfr.predict(x_test_cat).round(), 1, 81)
AAR, MAE, *_, sigmas, maes = aar(y_test, out)
ae = np.abs(y_test - out)
final_aar.append(AAR)
final_MAE.append(ae.mean())
####### refine rfc #########
t0 = time.time()
rrfc = RefinedRandomForest(rfc, C = 0.01, n_prunings = 1)
print(f'Time it took for refinement of rfc: {time.time() - t0:.3f} ms.')
leaves.append(sum(rrfc.n_leaves_))
for i in range(100):
t0 = time.time()
rrfc.fit(x_train, y_train_group)
# dump(rrfc, f'data/tlrf_rrfc_n_prunings_1.joblib')
# rrfc = joblib.load('data/tlrf_rrfc_n_prunings_1.joblib')
outc = rrfc.predict_proba(x_test)
y_pred_group = outc.argmax(axis=1)
#print(f"After refined Classifier accuracy: {accuracy_score(y_test_group, y_pred_group):.3f}")
#print(f"Report: \n {classification_report(y_test_group, y_pred_group)}")
classify_accuracy.append(accuracy_score(y_test_group, y_pred_group))
r = classification_report(y_test_group, y_pred_group).split()
weighted_f1.append(r[57])
leaves.append(sum(rrfc.n_leaves_))
x_test_cat = np.concatenate([x_test, outc], axis=1)
out = np.clip(rfr.predict(x_test_cat).round(), 1, 81)
AAR, MAE, *_, sigmas, maes = aar(y_test, out)
ae = np.abs(y_test - out)
final_aar.append(AAR)
final_MAE.append(ae.mean())
print(f'Time it took for pruning of rfc: {time.time() - t0:.3f} ms.')
####### refine rfr #########
# rfr = joblib.load('data/tlrf_resnext_rfr_100_5_128_1.joblib')
# t0 = time.time()
# rrfr = RefinedRandomForest(rfr, C = 0.01, n_prunings = 1)
# rrfr.fit(x_train, y_train)
# dump(rrfr, f'data/tlrf_rrfr_n_prunings_1.joblib')
# print(f'Time it took for refinement of rfr: {time.time() - t0:.3f} ms.')
# out = rrfr.predict_proba(x_test).argmax(axis=1)
# print(f'rrf MAE on validation: {np.abs(out - y_test).mean()}')
# ARR, *_ = aar(y_test, out)
# print(f'rrf AAR on validation: {ARR}')
print('classify_accuracy: ', classify_accuracy, '\n',
'weighted_f1', weighted_f1, '\n',
'final_aar', final_aar, '\n',
'final_MAE', final_MAE, '\n')
nprunings = np.arange(len(classify_accuracy))
fig, ax = plt.subplots(5, figsize=(14,11))
ax[0].plot(nprunings, classify_accuracy, '-o')
ax[0].set_title('Test set classify_accuracy')
ax[0].grid()
ax[1].plot(nprunings, weighted_f1, '-o')
ax[1].set_title('Test set weighted_f1 score')
ax[1].grid()
ax[2].plot(nprunings, final_aar, '-o')
ax[2].set_title('Test set AAR')
ax[2].grid()
ax[3].plot(nprunings, final_MAE, '-o')
ax[3].set_title('Test set MAE')
ax[3].grid()
ax[4].plot(nprunings, np.array(leaves)/1000, '-o')
ax[4].set_title('Number of leaves, thousands')
ax[4].set_xlabel('Number of prunings')
ax[4].grid()
#fig.savefig('ClassificationRefinement.png')