-
Notifications
You must be signed in to change notification settings - Fork 0
/
r4_evaluateSonoNERF.m
140 lines (103 loc) · 5.51 KB
/
r4_evaluateSonoNERF.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
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
%% Init:
clear
close all
clc
path(pathdef);
addpath( genpath( 'Source' ) )
load('DataCalculated/TrainedSonoNERFs/sonoNerfTrained - UALogo.mat')
%% Inference: preparation
frequenciesERTF = structSonoNERF.SonoNERF.Layers(2).frequencySamples(1:end/2);
numFrequencies = length( frequenciesERTF );
directionsERTF = structSonoNERF.SonoNERF.Layers(2).directionsSampling;
%% Inference: Spectrogram
dataTest = structSonoNERF.structSonoNERFData.dataChunkMatrix{104};
specgramPred = structSonoNERF.SonoNERF.predict( dataTest.inputData' )';
figure;
subplot(1,2,1 )
imagesc(specgramPred)
title('SonoNERF Prediction' )
subplot(1,2,2 )
imagesc(dataTest.outputData / structSonoNERF.scalerOutput)
title('Ground Truth' )
%% Inference: ERTF Extraction
[ azERTFSim, elERTFSim, ~ ] = cart2sph( directionsERTF(1,:), directionsERTF(2,:), directionsERTF(3,:) );
azERTFSim = rad2deg( azERTFSim );
elERTFSim = rad2deg( elERTFSim );
[ azMesh, elMesh ] = meshgrid( [-90:1:90 ], [-90:1:90 ] );
fullERTFSim = sqrt( structSonoNERF.SonoNERF.Layers(2).ERTFReal.^2 + structSonoNERF.SonoNERF.Layers(2).ERTFImag.^2 );
storageERTF = zeros( [ size(azMesh) numFrequencies 2 ] );
for cntFreq = 1 : numFrequencies
curSliceLeft = ( double( gather( fullERTFSim(:, cntFreq ) ) ) );
curSliceRight = ( double( gather( fullERTFSim(:, cntFreq + numFrequencies ) ) ) );
if( isa( curSliceLeft, 'dlarray' ) )
curSliceLeft = extractdata( curSliceLeft );
curSliceRight = extractdata( curSliceRight );
end
interpolatorERTFLeft = scatteredInterpolant( azERTFSim(:), elERTFSim(:), curSliceLeft(:) );
interpolatorERTFRight = scatteredInterpolant( azERTFSim(:), elERTFSim(:), curSliceRight(:) );
storageERTF( :,:,cntFreq,1) = interpolatorERTFLeft( azMesh, elMesh );
storageERTF( :,:,cntFreq,2) = interpolatorERTFRight( azMesh, elMesh );
end
figure;
idxFreqDisp = round( linspace( 1, 47, 4 ) );
for cntFreqDisp = 1 : 4
plotIdx = (cntFreqDisp-1)*2 + 1;
subplot(4,2,plotIdx)
imagesc( squeeze( storageERTF(:,:,idxFreqDisp(cntFreqDisp),1)))
subplot(4,2,plotIdx+1)
imagesc( squeeze( storageERTF(:,:,idxFreqDisp(cntFreqDisp),2)))
end
%% Inference: Extract IsoSurface
xVec = -0.3 : 0.005: 0.3;
yVec = -0.1 : 0.005 : 0.1;
zVec = -0.2 : 0.005 : 0.2;
[ xGrid, yGrid, zGrid ] = meshgrid( yVec, xVec, zVec );
% directionsInterrogation = structSonoNERF.SonoNERF.Layers(2).directionsSampling( :, 1 : 10 : end );
directionsInterrogation = eq_point_set( 2, 100 );
[ azInterrogation, elInterrogation, rInterrogation ] = cart2sph( directionsInterrogation(1,:), directionsInterrogation(2,:), directionsInterrogation(3,:) );
% azInterrogationVec = deg2rad( [ 0 -20 20 ] );
% elInterrogationVec = deg2rad( [ 0 -45 45 ] );
%
% [ azInterrogation, elInterrogation ] = meshgrid( azInterrogationVec, elInterrogationVec );
% azInterrogation = azInterrogation(:);
% elInterrogation = elInterrogation(:);
idxFreqCombiner = [ 1 : 10 ; 11 : 20 ; 21 : 30 ; 31 : 40 ];
dataStorageMatrix = zeros( length(xGrid(:)), length( azInterrogation ) );
for cntDirection = 1 : length( azInterrogation )
cntDirection
curAz = azInterrogation( cntDirection );
curEl = elInterrogation( cntDirection );
dataIn = gpuArray( [ xGrid(:) yGrid(:) zGrid(:) ones( size(zGrid(:) ) )*curEl ones( size(zGrid(:) ) )*curAz ] );
volumePredictedVecRaw = gather( structSonoNERF.SonoNERF.Layers(2).modelSonoNerf.predict(dataIn) );
% volumePredictedVecRaw = gather( net.Layers(2).modelSonoNerf.predict(dataIn) );
volumePredictedVecComplex = volumePredictedVecRaw(:,1:end/2) + 1i*volumePredictedVecRaw(:,end/2+1:end);
% volumePredictedVecEnergy = sqrt( sum( abs(volumePredictedVecComplex), 2 ) );
% volumePredictedVecEnergy = sum( abs(volumePredictedVecComplex(:, idxFreqCombiner(4,:)) ), 2 );
volumePredictedVecEnergy = sum( abs(volumePredictedVecComplex(:, :) ), 2 );
dataStorageMatrix( :, cntDirection ) = volumePredictedVecEnergy;
end
%%
% volumePredictedVecEnergy = sqrt( sum( ( dataStorageMatrix.^2 ), 2 ) );
% volumePredictedVecEnergy = prod( dataStorageMatrix / max(dataStorageMatrix(:)), 2 );
% volumePredictedVecEnergy = max( dataStorageMatrix, [], 2 );
volumePredictedVecEnergy = sum( dataStorageMatrix, 2 );
volumePredictedEnergy = reshape( volumePredictedVecEnergy, [ length(xVec) length(yVec) length(zVec)] );
volumePredictedEnergy = smooth3( volumePredictedEnergy, 'box', 9 );
volshow( volumePredictedEnergy )
%
% thresholdEstimator = mode( xGrid(:) ) *300;
% % thresholdEstimator = 15
% [faces,verts] = isosurface(xGrid, yGrid, zGrid, volumePredictedEnergy, thresholdEstimator);
%
% surfaceMeshExtracted = surfaceMesh( verts, faces );
% surfaceMeshSmooth = smoothSurfaceMesh(surfaceMeshExtracted,5);
% surfaceMeshShow(surfaceMeshSmooth,Title="Extracted SonoNERF Surface")
%
% % figure;
% p = patch('faces',surfaceMeshSmooth.Faces,'vertices', surfaceMeshSmooth.Vertices, 'FaceColor', [ 0.2 0.3 0.3 ], 'EdgeAlpha', 0.3);
% axis equal
% grid on
% camlight
% xlabel( 'X-axis (m)' )
% ylabel( 'Y-axis (m)' )
% zlabel( 'Z-axis (m)' )