Graphics Reference
In-Depth Information
Table 4.1
RMS errors (mm) of output models to input scans
Data sets
model
min
max
mean
sd
UND
1 comp
0.51
1.73
0.79
0.13
4 comp
0.39
1.57
0.66
0.11
7 comp
0.39
1.50
0.64
0.10
GAVAB
1 comp
0.94
3.45
1.22
0.19
4 comp
0.80
2.30
1.06
0.14
7 comp
0.80
2.24
1.05
0.14
BU-3DFE
1 comp
0.92
1.97
1.20
0.18
4 comp
0.87
1.77
1.09
0.15
7 comp
0.87
1.78
1.08
0.16
4.4 Dynamic Model Expansion
For a statistical face model to be applicable in face recognition systems all over the world, it is
important to include example data on all possible face variations. Because this is an intractable
task, a flexible model is required that updates itself in case of new example faces. For that,
a system should automatically fit the face model to face data, estimate dense and accurate
correspondences beyond the linear combinations of current example data, and measure the
redundancy of the new example faces.
Table 4.2
Projection errors (mm) of output models to input scans
Data sets
model
min
max
mean
sd
UND
1 comp
0.40
1.65
0.65
0.15
4 comp
0.28
1.33
0.47
0.11
7 comp
0.26
1.20
0.43
0.10
GAVAB
1 comp
0.49
3.03
0.79
0.20
4 comp
0.35
1.91
0.55
0.14
7 comp
0.35
1.69
0.53
0.12
BU-3DFE
1 comp
0.37
1.58
0.63
0.17
4 comp
0.25
0.99
0.43
0.10
7 comp
0.23
0.92
0.39
0.11
