-
Notifications
You must be signed in to change notification settings - Fork 1
/
runme.m
48 lines (38 loc) · 1.38 KB
/
runme.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
%% Important!
%
% The training process takes a long time if run on a single machine.
% You could look at qsub_run_xxx.sh scripts for parallel computing.
% Contact Ning Ma ([email protected]) for any problem.
%
startTwoEars('Config.xml');
%% ===== Configuration =====
%
% Need to set up relevant paths in get_data_root.m
% dataRoot - default root for storing features etc
% twoearsRoot - TwoEars git root
% Training parameters
preset = 'MCT-DIFFUSE'; % Multiconditional training
azRes = 1; % Azimuth resolution in degrees
azimuthVector = 0:azRes:359; % Vector of azimuths considered
%% ===== Create training and development features =====
%
% Create raw features
f1_createBinauralFeatureTrain(azimuthVector, preset, azRes);
f1_createBinauralFeatureDev(azimuthVector, azRes);
%
% Process features (normalisation etc)
AFE_param = initialiseAfeParameters();
channelVector = 1:AFE_param.fb_nChannels;
f2_processBinauralFeatureTrain(channelVector, preset, azRes);
f2_processBinauralFeatureDev(channelVector, preset, azRes);
%% ===== Train models =====
%
% Use 'itd-ild' for training GmmLocationKS
featureType = 'itd-ild';
f3_trainGmmLocationKS(preset, featureType, azRes);
% Use 'ild-cc' for training DnnLocationKS
featureType = 'ild-cc';
for channel = channelVector
f3_trainDnnLocationKS(channel, preset, featureType, azRes);
end
% vim: set sw=4 ts=4 et tw=90: