/
metrics.py
78 lines (57 loc) · 2.48 KB
/
metrics.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
"""
Adapted from https://github.com/zhufengx/SRN_multilabel
"""
import numpy as np
def calculate_metrics(labels, preds):
mAP = calculate_mAP(labels, preds)
pc_top3, rc_top3, f1c_top3, po_top3, ro_top3, f1o_top3 = calculate_top3_metrics(labels, preds)
return {'pc_top3': pc_top3, 'rc_top3': rc_top3, 'f1c_top3': f1c_top3,
'po_top3': po_top3, 'ro_top3': ro_top3, 'f1o_top3': f1o_top3, 'mAP': mAP}
def calculate_top3_metrics(labels, preds):
no_examples = labels.shape[0]
top3 = np.zeros_like(preds)
for ind_example in range(no_examples):
top_pred_inds = np.argsort(preds[ind_example])[::-1]
for k in range(3):
top3[ind_example, top_pred_inds[k]] = 1
pc_top3, rc_top3, f1c_top3 = prec_rec_f1(labels, top3)
po_top3, ro_top3, f1o_top3 = prec_rec_f1(labels.flatten(), top3.flatten())
return pc_top3, rc_top3, f1c_top3, po_top3, ro_top3, f1o_top3
def prec_rec_f1(labels, pred_labels):
eps = np.finfo(np.float32).eps
tp = labels * pred_labels
if len(labels.shape) == 2:
no_tp = np.sum(tp, axis=1) + eps
no_pred = np.sum(pred_labels, axis=1) + eps
no_pos = np.sum(labels, axis=1) + eps
elif len(labels.shape) == 1:
no_tp = np.sum(tp) + eps
no_pred = np.sum(pred_labels) + eps
no_pos = np.sum(labels) + eps
prec_class = no_tp / no_pred + eps
rec_class = no_tp / no_pos + eps
f1_class = 2 * prec_class * rec_class / (prec_class + rec_class)
return 100 * np.mean(prec_class), 100 * np.mean(rec_class), 100 * np.mean(f1_class)
def calculate_mAP(labels, preds):
no_examples = labels.shape[0]
no_classes = labels.shape[1]
ap_scores = np.empty((no_classes), dtype=np.float)
for ind_class in range(no_classes):
ground_truth = labels[:, ind_class]
out = preds[:, ind_class]
sorted_inds = np.argsort(out)[::-1] # in descending order
tp = ground_truth[sorted_inds]
fp = 1 - ground_truth[sorted_inds]
tp = np.cumsum(tp)
fp = np.cumsum(fp)
rec = tp / np.sum(ground_truth)
prec = tp / (fp + tp)
rec = np.insert(rec, 0, 0)
rec = np.append(rec, 1)
prec = np.insert(prec, 0, 0)
prec = np.append(prec, 0)
for ind in range(no_examples, -1, -1):
prec[ind] = max(prec[ind], prec[ind + 1])
inds = np.where(rec[1:] != rec[:-1])[0] + 1
ap_scores[ind_class] = np.sum((rec[inds] - rec[inds - 1]) * prec[inds])
return 100 * np.mean(ap_scores)