-
Notifications
You must be signed in to change notification settings - Fork 5
/
config_aff_ori_desc_zeromq.ini
128 lines (128 loc) · 5.82 KB
/
config_aff_ori_desc_zeromq.ini
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
[Computing]
LoadColor=1;
numberOfCores=-1; how many cores to use. -1 for maximum possible
[AffineAdaptation]
useZMQ=1
[AffNet]
mrSize=5.1962;
port=tcp://localhost:5556
patchSize=32; width and height of the patch
[DominantOrientation]
;patch size and mrSize here are use if useZMQ=false. If useZMQ=1, ones from [OriNet] are used
mrSize = 5.1962; = 3.0f*sqrt(3.0f). size of the measurement region (as multiple of the feature scale) for orientation estimation.
patchSize = 32; width and height of the patch
maxAngles = 1
threshold = 0.8; fraction of max peak to add peak
addUpright=false
halfSIFTMode=false;
useZMQ=1
addMirrored=false
[OriNet]
port=tcp://localhost:5557
patchSize=32; width and height of the patch
mrSize=5.1962; = 3.0f*sqrt(3.0f). size of the measurement region (as multiple of the feature scale) for orientation estimation
[zmqDescriptor]
port=tcp://localhost:5555
patchSize=32; width and height of the patch
mrSize=5.1962 ; = 3.0f*sqrt(3.0f). size of the measurement region (as multiple of the feature scale) for SIFT computation
[MSER] ;CMP implementation
max_area=0.05; maximum area of the MSER to the image area
min_size=30 ; minimum region size (in pixels)
min_margin=8; threshold
mode=FixedTh;RegNumber;can be "FixedTh", "FixedRegNumber", "RelativeTh", RelativeRegNumber
;relativeThreshold=0.0001;
regionsNumber = 500;
;relativeRegionsNumber = 0.2;
[HessianAffine];CMP implementation
mode=FixedTh; "FixedRegNumber", "RelativeTh", RelativeRegNumber, NotLessThanRegions
threshold=5.33;18.0;5.333; = 16.0f/3.0f // noise dependent threshold on the response (sensitivity)
;relativeThreshold=0.1;
regionsNumber=3000;
;relativeRegionsNumber = 0.5;
max_iter = 16; number of affine shape interations
numberOfScales= 3; number of scale per octave
initialSigma = 1.6; amount of smoothing applied to the initial level of first octave
edgeEigenValueRatio = 10.0; ratio of the eigenvalues
border = 5; number of pixels ignored at the border of image
smmWindowSize = 19; // width and height of the SMM mask, pixels
convergenceThreshold=0.05; //for Baumberg iteration
doBaumberg=0; Mean "do internal affine shape adaptation. If you want to use external, set AffineAdaptation=0 here and useZMQ=1 in [AffineAdaptation].
sampleFromImage=0
method=SMM
[HarrisAffine];CMP implementation
mode=FixedTh;NotLessThanRegions;FixedRegNumber;can be "FixedTh", "FixedRegNumber", "RelativeTh", RelativeRegNumber
threshold = 15.0; = 16.0f/3.0f // noise dependent threshold on the response (sensitivity)
relativeThreshold=0.1;
regionsNumber = 1000;
relativeRegionsNumber = 0.5;
max_iter = 16; number of affine shape interations
numberOfScales= 3; number of scale per octave
initialSigma = 1.6; amount of smoothing applied to the initial level of first octave
edgeEigenValueRatio = 10.0; ratio of the eigenvalues
border = 5; number of pixels ignored at the border of image
smmWindowSize = 19; // width and height of the SMM mask, pixels
convergenceThreshold=0.1; //for Baumberg iteration
HarrisCoef = 0.04;
doBaumberg=0; Mean "do internal affine shape adaptation. If you want to use external, set AffineAdaptation=0 here and useZMQ=1 in [AffineAdaptation].
[DoG];CMP implementation
mode=FixedTh;RegNumber; can be "FixedTh", "FixedRegNumber", "RelativeTh", RelativeRegNumber
threshold = 8.0; = 16.0f/3.0f // noise dependent threshold on the response (sensitivity)
relativeThreshold=0.01;
regionsNumber = 3000;
relativeRegionsNumber = 0.5;
numberOfScales = 3; number of scale per octave
initialSigma = 1.6; amount of smoothing applied to the initial level of first octave
edgeEigenValueRatio =10.0; ratio of the eigenvalues
border = 5; number of pixels ignored at the border of image
max_iter = 32; number of affine shape interations
border = 5; number of pixels ignored at the border of image
convergenceThreshold=0.05; //for Baumberg iteration
smmWindowSize = 19; // width and height of the SMM mask, pixels
doBaumberg=0; Mean "do internal affine shape adaptation. If you want to use external, set AffineAdaptation=0 here and useZMQ=1 in [AffineAdaptation].
[SIFTDescriptor] ;CMP implementation
photoNorm=true
FastPatchExtraction=false
spatialBins = 4; width and height bins
orientationBins = 8;
maxBinValue = 0.2;0.2;
patchSize = 41; width and height of the patch
mrSize = 5.1962 ; = 3.0f*sqrt(3.0f). size of the measurement region (as multiple of the feature scale) for SIFT computation
[Matching]; FLANN from OpenCV
binary_matcher=linear; Can be linear, lsh, hierarchical.
vector_matcher=kdtree;kdtree;Can be linear, kdtree,composite, kmeans
binary_dist=Hamming; Can be Hamming
vector_dist=L2;Can be L1,L2
doBothRANSACgroundTruth = 1; only when used ground truth mode
RANSACforStopping = 1; only when used ground truth mode and doBothRANSACgroundTruth =1
contradDist = 10.0;10.0; in pixels
checkFwithPTS = 0
kd_trees=4;
knn_checks=128;
[DuplicateFiltering]
doBeforeRANSAC = 1 ;if == 0, tend to degenerated cases faster but a bit less true mathces
duplicateDist = 2.0 ;in pixels
whichCorrespondenceRemains=bestFGINN;can random, bestFGINN, bestDistance, biggerRegion
[RANSAC] ;CMP implementation
err_threshold = 4.0;in pixels.
ErrorType=Sampson; Can be Sampson,SymmSum.
doSymmCheck = 1;
confidence = 0.99
max_samples = 1000000
localOptimization = 1
LAFcoef=2.0; LAFcoef*err_threshold = threshold for LAF-F-check. Use 0 to switch check off
HLAFcoef=12.0; LAFcoef*err_threshold = threshold for Symmetrical LAF-H-check. Use 0 to switch check off
;Symmetrical error >= 2* Sampson error, also LAF can be not precisely equal. 12 is recommended
[ImageOutput]
writeImages = 1
drawOnlyCenters = 0; if 0, all matches ellipses with orientations will be shown.
drawEpipolarLines = 0; in F-mode only.
drawReprojected = 0;0 == ASIFT-like output
[TextOutput]
verbose=1;
timeLog= 1;
featureComplemetaryLog = 0;
writeKeypoints = 1;
writeMatches = 0;
outputAllTentatives=0;
outputEstimatedHorF=0;
outputMikFormat=1; 1 means x y a b c desc[], 0 means x y s a11 a12 a21 a22 desc[]