-
Notifications
You must be signed in to change notification settings - Fork 326
/
evaluation.m
44 lines (36 loc) · 1.44 KB
/
evaluation.m
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
% load data
% load('../results/LightenedCNN_A_lfw.mat'); % model A
% load('../results/LightenedCNN_B_lfw.mat'); % model B
load('../results/LightenedCNN_C_lfw.mat'); % model C
load('lfw_pairs.mat');
% pos
for i = 1: length(pos_pair)
feat1 = features(pos_pair(1, i), :)';
feat2 = features(pos_pair(2, i), :)';
pos_scores(i) = distance.compute_cosine_score(feat1, feat2);
% pos_scores(i) = -distance.compute_L2_score(feat1, feat2);
end
pos_label = ones(1, length(pos_pair));
%neg
for i = 1: length(neg_pair)
feat1 = features(neg_pair(1, i), :)';
feat2 = features(neg_pair(2, i), :)';
neg_scores(i) = distance.compute_cosine_score(feat1, feat2);
% neg_scores(i) = -distance.compute_L2_score(feat1, feat2);
end
neg_label = -ones(1, length(neg_pair));
scores = [pos_scores, neg_scores];
label = [pos_label neg_label];
% ap
ap = evaluation.evaluate('ap', scores, label);
% roc
roc = evaluation.evaluate('roc', scores, label);
%% output
fprintf('ap: %f\n', ap.measure);
fprintf('eer: %f\n', roc.measure);
fprintf('tpr001: %f\n', roc.extra.tpr001*100);
fprintf('tpr0001: %f\n', roc.extra.tpr0001*100);
fprintf('tpr00001: %f\n', roc.extra.tpr00001*100);
fprintf('tpr000001: %f\n', roc.extra.tpr000001*100);
fprintf('tpr0: %f\n', roc.extra.tpr0*100);
result = [ap.measure/100 roc.measure/100 roc.extra.tpr001 roc.extra.tpr0001 roc.extra.tpr00001 roc.extra.tpr000001 roc.extra.tpr0];