text
stringlengths 0
89.3k
|
|---|
2Refer to more analysis on proxy shape and its role in the Supp Sec 3Algorithm 1 Overview of the Optimization Algorithm |
1Input Initialize SSˆSSvia Pseudo inverse shape |
in 11 RpQIΛIβ 1e4andϵ 1e6 |
2repeat |
3 Updating variables SˆSSSalternately accord |
ing to model 13 yields SˆSSS |
4untilSS ϵ |
5Initialization Calculate Arby Eq 9 build weight |
matrix Λby 10 reset ββd |
6repeat |
7 Updating variables Ωalternately according to model |
13 yields SˆSSSQ |
8untilSS ϵ |
Output SQ |
TPA module 4 The complete model is as follows |
min |
SQµ1 |
2WΠS2 |
Fµ2 |
S |
µ3 |
2F1X |
i1 |
QiSiQi1Si1 |
2 |
F |
st |
SgˆSΛ |
ˆSiQiSi i 1 F |
SiRpiSi i 1 F |
12 |
where ΠdiagΠiΠiR23is the camera projec |
tion matrix and we use orthogonal projection as 11 26 |
SR3FPis the 3D shape sequence under the camera |
coordinate Then we improve the initialized camera motion |
Rpby estimating QiSO3and compute the 3D shape |
sequence ˆSR3FPin the canonical coordinate Finally |
we introduce the kernel based weight matrix Λto construct |
proxy shapes that more satisfy the lowrank constraint |
We optimize the model 12 by ADMM 6 and the La |
grange multiplier of the whole model is as follows |
min |
ΩLµ1 |
2WΠS2 |
Fµ2 |
S |
µ3 |
2F1X |
i1 |
QiSiQi1Si1 |
2 |
F |
β |
2 |
SgˆSΛ |
2 |
FD |
Y1SgˆSΛE |
β |
2 |
ˆSQS |
2 |
FD |
Y2ˆSQSE |
β |
2 |
SRpS |
2 |
FD |
Y3SRpSE13 |
where Ω SSˆSSQdenotes the variables to be up |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.