/
test_local_novelty_detection_model.m
100 lines (71 loc) · 3.88 KB
/
test_local_novelty_detection_model.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
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
% Testing method for local novelty detection according to the work:
%
% Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler:
% "Local Novelty Detection in Multi-class Recognition Problems".
% Proceedings of the IEEE Winter Conference on Applications of Computer Vision (WACV), 2015.
%
% Please cite that paper if you are using this code!
%
%
% function scores = test_local_novelty_detection_model( model, Ks )
%
% Calculate novelty scores following local novelty detection and using the model structure obtained from learn_local_novelty_detection_model.
% For each test sample, a novelty detection model is learned on-the-fly using its k nearest neighbors of the training set.
% Models are either one-class or multi-class models depending on the class distribution in the neighborhood
%
% INPUT:
% model -- model obtained from learn_local_novelty_detection_model
% Ks -- (n x m) kernel matrix containing similarities between n training samples and m test samples
%
% OUTPUT:
% scores -- novelty scores for the m test samples obtained by exemplar-specific local novelty detection models
%
% (LGPL) copyright by Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler
%
function scores = test_local_novelty_detection_model( model, Ks )
%%% get number of test samples
n_test_samples = size(Ks,2);
%%% init scores
scores = zeros(n_test_samples, 1);
%%% check size of Ks
if size(Ks,1) ~= size(model.K,1)
error('number of rows of input "Ks" must match number of rows of "model.K"');
end
%%% loop over test samples
for t=1:n_test_samples
%%% sort train samples according to the distance to the current test sample in the kernel feature space
[~,sort_id] = sort(Ks(:,t),'descend');
%%% get IDs of the nearest neighbors
sort_id = sort_id(1:min(model.numNeighbors,length(model.labels)));
%%% check whether all neighbors belong to the same class
occ_setting = length(unique(model.labels(sort_id))) == 1;
%%% check for OCC setting
if occ_setting
%%% ================================
%%% One-class classification setting
%%% ================================
if ~isempty(model.methods_params{1})
%%% learn a local one-class model with additional method parameters taking only the nearest neighbors into account
local_model = feval(sprintf('learn_%s',model.methods{1}) , model.K( sort_id , sort_id ) , model.methods_params{1}{:} );
else
%%% learn a local one-class model without additional method parameters (such as KNFST) taking only the nearest neighbors into account
local_model = feval(sprintf('learn_%s',model.methods{1}) , model.K( sort_id , sort_id ) );
end
%%% evaluate the local one-class model for the current test sample
scores(t) = feval(sprintf('test_%s',model.methods{1}) , local_model , Ks( sort_id , t ) );
else
%%% =====================================
%%% Multi-class novelty detection setting
%%% =====================================
if ~isempty(model.methods_params{2})
%%% learn a local multi-class model with additional method parameters taking only the nearest neighbors into account
local_model = feval(sprintf('learn_%s',model.methods{2}) ,model.K( sort_id , sort_id ) , model.labels(sort_id), model.methods_params{2}{:} );
else
%%% learn a local multi-class model without additional method parameters (such as KNFST) taking only the nearest neighbors into account
local_model = feval(sprintf('learn_%s',model.methods{2}) ,model.K( sort_id , sort_id ) , model.labels(sort_id) );
end
%%% evaluate the local multi-class model for the current test sample
scores(t) = feval(sprintf('test_%s',model.methods{2}) , local_model , Ks( sort_id , t ) );
end
end
end