forked from potterhsu/SVHNClassifier-PyTorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluator.py
147 lines (123 loc) · 6.38 KB
/
evaluator.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
import torch
import torch.utils.data
from sklearn.metrics import (
accuracy_score,
classification_report,
confusion_matrix,
f1_score,
precision_score,
recall_score,
roc_auc_score,
)
from torchvision import transforms
from .dataset import Dataset
class Evaluator(object):
def __init__(self, path_to_lmdb_dir):
transform = transforms.Compose([
transforms.CenterCrop([54, 54]),
transforms.ToTensor(),
transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
self._loader = torch.utils.data.DataLoader(Dataset(path_to_lmdb_dir, transform), batch_size=32, shuffle=False)
def get_model_metrics(self, model):
y_pred = []
y_true = []
with torch.no_grad():
for batch_idx, (images, length_labels, digits_labels, _) in enumerate(self._loader):
images, length_labels, digits_labels = images.cuda(), length_labels.cuda(), [digit_labels.cuda() for digit_labels in digits_labels]
length_logits, digit1_logits, digit2_logits, digit3_logits, digit4_logits, digit5_logits = model.eval()(images)
# Predictions
digit1_prediction = digit1_logits.max(1)[1]
digit2_prediction = digit2_logits.max(1)[1]
digit3_prediction = digit3_logits.max(1)[1]
digit4_prediction = digit4_logits.max(1)[1]
digit5_prediction = digit5_logits.max(1)[1]
predictions = [
digit1_prediction,
digit2_prediction,
digit3_prediction,
digit4_prediction,
digit5_prediction,
]
for prediction, label in zip(predictions, digits_labels):
for digit_prediction, digit_label in zip(prediction, label):
y_pred.append(str(int(digit_prediction)))
y_true.append(str(int(digit_label)))
# Specificity and Sensitivity
elements = {}
for true, prediction in zip(y_true, y_pred):
if true not in elements:
elements[true] = {"total": 0, "true_positives": 0, "false_positives": 0}
if prediction not in elements:
elements[prediction] = {"total": 0, "true_positives": 0, "false_positives": 0}
elements[true]["total"] += 1
if prediction == true:
elements[true]["true_positives"] += 1
else:
elements[prediction]["false_positives"] += 1
for digit in elements:
elements[digit]["sensitivity"] = elements[digit]["true_positives"] / elements[digit]["total"]
total_negatives = len(y_true) - elements[digit]["total"]
elements[digit]["specificity"] = (total_negatives - elements[digit]["false_positives"]) / total_negatives
accuracy = accuracy_score(y_true, y_pred)
f1 = f1_score(y_true, y_pred, average="weighted")
precision = precision_score(y_true, y_pred, average="weighted")
recall = recall_score(y_true, y_pred, average="weighted")
matrix = confusion_matrix(y_true, y_pred,
labels=["0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "10"])
report = classification_report(y_true, y_pred, output_dict=True, digits=2)
# Round report to 2 decimal places
for label, metrics in report.items():
if label in elements:
report[label]["specificity"] = elements[label]["specificity"]
report[label]["sensitivity"] = elements[label]["sensitivity"]
if label != "accuracy":
for metric, value in metrics.items():
metrics[metric] = round(value, 2)
# Add row/column totals to confusion matrix
matrix = matrix.tolist()
col_sums = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
for row in matrix:
row.append(sum(row))
for idx, col in enumerate(row):
col_sums[idx] += col
matrix.append(col_sums)
model_info = {
"accuracy": round(accuracy, 2),
"confusion_matrix": matrix,
"classification_report": report,
"precision": round(precision, 2),
"f1_score": round(f1, 2),
"recall": round(recall, 2),
}
return model_info
def evaluate(self, model):
num_correct = 0
needs_include_length = False
with torch.no_grad():
for batch_idx, (images, length_labels, digits_labels, _) in enumerate(self._loader):
images, length_labels, digits_labels = images.cuda(), length_labels.cuda(), [digit_labels.cuda() for digit_labels in digits_labels]
length_logits, digit1_logits, digit2_logits, digit3_logits, digit4_logits, digit5_logits = model.eval()(images)
if batch_idx == 0:
details = [length_logits, digit1_logits, digit2_logits, digit3_logits, digit4_logits, digit5_logits, length_labels, digits_labels]
length_prediction = length_logits.max(1)[1]
digit1_prediction = digit1_logits.max(1)[1]
digit2_prediction = digit2_logits.max(1)[1]
digit3_prediction = digit3_logits.max(1)[1]
digit4_prediction = digit4_logits.max(1)[1]
digit5_prediction = digit5_logits.max(1)[1]
if needs_include_length:
num_correct += (length_prediction.eq(length_labels) &
digit1_prediction.eq(digits_labels[0]) &
digit2_prediction.eq(digits_labels[1]) &
digit3_prediction.eq(digits_labels[2]) &
digit4_prediction.eq(digits_labels[3]) &
digit5_prediction.eq(digits_labels[4])).cuda().sum()
else:
num_correct += (digit1_prediction.eq(digits_labels[0]) &
digit2_prediction.eq(digits_labels[1]) &
digit3_prediction.eq(digits_labels[2]) &
digit4_prediction.eq(digits_labels[3]) &
digit5_prediction.eq(digits_labels[4])).cuda().sum()
accuracy = num_correct.item() / len(self._loader.dataset)
return accuracy, details