368 lines
15 KiB
TeX
368 lines
15 KiB
TeX
\documentclass[portrait,final,a0paper]{baposter}
|
|
%\documentclass[a4shrink,portrait,final]{baposter}
|
|
% Usa a4shrink for an a4 sized paper.
|
|
|
|
\tracingstats=2
|
|
|
|
\usepackage{calc}
|
|
\usepackage{graphicx}
|
|
\usepackage{amsmath}
|
|
\usepackage{amssymb}
|
|
\usepackage{relsize}
|
|
\usepackage{multirow}
|
|
\usepackage{bm}
|
|
|
|
\usepackage{graphicx}
|
|
\usepackage{multicol}
|
|
|
|
\usepackage{pgfbaselayers}
|
|
\pgfdeclarelayer{background}
|
|
\pgfdeclarelayer{foreground}
|
|
\pgfsetlayers{background,main,foreground}
|
|
|
|
\usepackage{times}
|
|
\usepackage{helvet}
|
|
%\usepackage{bookman}
|
|
\usepackage{palatino}
|
|
|
|
\newcommand{\captionfont}{\footnotesize}
|
|
|
|
\selectcolormodel{cmyk}
|
|
|
|
\graphicspath{{images/}}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%%% Some math symbols used in the text
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Format
|
|
\newcommand{\Matrix}[1]{\begin{bmatrix} #1 \end{bmatrix}}
|
|
\newcommand{\Vector}[1]{\Matrix{#1}}
|
|
\newcommand*{\SET}[1] {\ensuremath{\mathcal{#1}}}
|
|
\newcommand*{\MAT}[1] {\ensuremath{\mathbf{#1}}}
|
|
\newcommand*{\VEC}[1] {\ensuremath{\bm{#1}}}
|
|
\newcommand*{\CONST}[1]{\ensuremath{\mathit{#1}}}
|
|
\newcommand*{\norm}[1]{\mathopen\| #1 \mathclose\|}% use instead of $\|x\|$
|
|
\newcommand*{\abs}[1]{\mathopen| #1 \mathclose|}% use instead of $\|x\|$
|
|
\newcommand*{\absLR}[1]{\left| #1 \right|}% use instead of $\|x\|$
|
|
|
|
\def\norm#1{\mathopen\| #1 \mathclose\|}% use instead of $\|x\|$
|
|
\newcommand{\normLR}[1]{\left\| #1 \right\|}% use instead of $\|x\|$
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Multicol Settings
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\setlength{\columnsep}{0.7em}
|
|
\setlength{\columnseprule}{0mm}
|
|
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Save space in lists. Use this after the opening of the list
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\newcommand{\compresslist}{%
|
|
\setlength{\itemsep}{1pt}%
|
|
\setlength{\parskip}{0pt}%
|
|
\setlength{\parsep}{0pt}%
|
|
}
|
|
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%% Begin of Document
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
|
|
\begin{document}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%% Here starts the poster
|
|
%%%---------------------------------------------------------------------------
|
|
%%% Format it to your taste with the options
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
% Define some colors
|
|
\definecolor{silver}{cmyk}{0,0,0,0.3}
|
|
\definecolor{yellow}{cmyk}{0,0,0.9,0.0}
|
|
\definecolor{reddishyellow}{cmyk}{0,0.22,1.0,0.0}
|
|
\definecolor{black}{cmyk}{0,0,0.0,1.0}
|
|
\definecolor{darkYellow}{cmyk}{0,0,1.0,0.5}
|
|
\definecolor{darkSilver}{cmyk}{0,0,0,0.1}
|
|
|
|
\definecolor{lightyellow}{cmyk}{0,0,0.3,0.0}
|
|
\definecolor{lighteryellow}{cmyk}{0,0,0.1,0.0}
|
|
\definecolor{lighteryellow}{cmyk}{0,0,0.1,0.0}
|
|
\definecolor{lightestyellow}{cmyk}{0,0,0.05,0.0}
|
|
|
|
%%
|
|
\typeout{Poster Starts}
|
|
\background{
|
|
\begin{tikzpicture}[remember picture,overlay]%
|
|
\draw (current page.north west)+(-2em,2em) node[anchor=north west] {\includegraphics[height=1.1\textheight]{silhouettes_background}};
|
|
\end{tikzpicture}%
|
|
}
|
|
|
|
\newlength{\leftimgwidth}
|
|
\begin{poster}%
|
|
% Poster Options
|
|
{
|
|
% Show grid to help with alignment
|
|
grid=false,
|
|
% Column spacing
|
|
colspacing=1em,
|
|
% Color style
|
|
bgColorOne=lighteryellow,
|
|
bgColorTwo=lightestyellow,
|
|
borderColor=reddishyellow,
|
|
headerColorOne=yellow,
|
|
headerColorTwo=reddishyellow,
|
|
headerFontColor=black,
|
|
boxColorOne=lightyellow,
|
|
boxColorTwo=lighteryellow,
|
|
% Format of textbox
|
|
textborder=roundedleft,
|
|
% textborder=rectangle,
|
|
% Format of text header
|
|
eyecatcher=false,
|
|
headerborder=open,
|
|
headerheight=0.08\textheight,
|
|
headershape=roundedright,
|
|
headershade=plain,
|
|
headerfont=\Large\textsf, %Sans Serif
|
|
boxshade=plain,
|
|
% background=shade-tb,
|
|
background=plain,
|
|
linewidth=2pt
|
|
}
|
|
% Eye Catcher
|
|
{\includegraphics[width=10em]{D1077}} % No eye catcher for this poster. (eyecatcher=no above). If an eye catcher is present, the title is centered between eye-catcher and logo.
|
|
% Title
|
|
{\sf %Sans Serif
|
|
%\bf% Serif
|
|
Expression Invariant Face Recognition with a 3DMM}
|
|
% Authors
|
|
{\sf %Sans Serif
|
|
% Serif
|
|
\vspace{1em}\{Brian.Amberg, Reinhard.Knothe, Thomas.Vetter\}@unibas.ch
|
|
}
|
|
% University logo
|
|
{% The makebox allows the title to flow into the logo, this is a hack because of the L shaped logo.
|
|
\makebox[8em][r]{%
|
|
\begin{minipage}{16em}
|
|
\hfill
|
|
\includegraphics[height=2em]{msrlogo}
|
|
\includegraphics[height=7.0em]{logo}
|
|
\end{minipage}
|
|
}
|
|
}
|
|
|
|
\tikzstyle{light shaded}=[top color=baposterBGtwo!30!white,bottom color=baposterBGone!30!white,shading=axis,shading angle=30]
|
|
|
|
% Width of left inset image
|
|
\setlength{\leftimgwidth}{0.78em+8.0em}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%%% Now define the boxes that make up the poster
|
|
%%%---------------------------------------------------------------------------
|
|
%%% Each box has a name and can be placed absolutely or relatively.
|
|
%%% The only inconvenience is that you can only specify a relative position
|
|
%%% towards an already declared box. So if you have a box attached to the
|
|
%%% bottom, one to the top and a third one which should be in between, you
|
|
%%% have to specify the top and bottom boxes before you specify the middle
|
|
%%% box.
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
%
|
|
% A coloured circle useful as a bullet with an adjustably strong filling
|
|
\newcommand{\colouredcircle}[1]{%
|
|
\tikz{\useasboundingbox (-0.2em,-0.32em) rectangle(0.2em,0.32em); \draw[draw=black,fill=baposterBGone!80!black!#1!white,line width=0.03em] (0,0) circle(0.18em);}}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Contribution}{name=contribution,column=0,row=0}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
{}We introduce a method for expression invariant face recognition. A
|
|
generative 3D Morphable Model (3DMM) is used to separate identity and
|
|
expression components. The expression removal results in increased
|
|
recognition performance, even on difficult datasets, without a decrease in
|
|
performance on expression-less datasets.
|
|
|
|
It is applicable to any kind of input data, and was evaluated here on
|
|
textureless range scans.
|
|
\vspace{0.3em}
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Model}{name=model,column=0,below=contribution}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
The Model was learnt from 275 subjects. We used one neutral expression scan
|
|
per identity and 150 expression scans of a subset of the subjects.
|
|
|
|
The identity model is a linear model build from the neutral scans.
|
|
\begin{align}
|
|
\VEC f&=\VEC\mu + \MAT M_n\VEC\alpha_n\qquad.
|
|
\end{align}
|
|
For each of the 150 expression scans, we calculated an expression vector as
|
|
the difference between the expression scan and the corresponding neutral
|
|
scan of that subject. This data is already mode-centered, if we regard the
|
|
neutral expression as the natural mode of expression data. From these offset
|
|
vectors an additional expression matrix $\MAT M_e$ was calculated, such that the complete linear Model is
|
|
\begin{align}
|
|
\VEC f&=\VEC\mu + \MAT M_n\VEC\alpha_n + \MAT M_e\VEC\alpha_e
|
|
\end{align}
|
|
The assumption here is, that the face and expression space are linearly
|
|
independent, such that each face is represented by a unique set of
|
|
coefficients.
|
|
\vspace{0.3em}
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Fitting}{name=fitting,column=0,below=model}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
A Robust Nonrigid ICP method was used to fit the model to the data.
|
|
Robustness was achieved by iteratively reweighting the correspondences and
|
|
using hard compatability test for the closest points.
|
|
|
|
Fitting was initialized by a simple nose detector and proceeded fully
|
|
automatic.
|
|
\vspace{0.3em}
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Distance Measure}{name=measure,column=0,below=fitting}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
The Mahalanobis angle between the identity coefficients $\VEC{\alpha_{n}}$
|
|
was used for classification.
|
|
\vspace{0.3em}
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Expression Neutralization}{name=results neutralization,column=1,row=0}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\begin{tikzpicture}[x=0.3333\linewidth,y=-0.42\linewidth]
|
|
\path [use as bounding box] (-0.5,-0.5) rectangle(2.5,1.7);
|
|
\path
|
|
(0,0) node{\includegraphics[width=0.42\linewidth]{D1077}}
|
|
(1,0) node{\includegraphics[width=0.47\linewidth]{D1077_fit_expression}}
|
|
(2,0) node{\includegraphics[width=0.47\linewidth]{D1077_fit}}
|
|
|
|
(0,1) node{\includegraphics[width=0.42\linewidth]{D1360}}
|
|
(1,1) node{\includegraphics[width=0.47\linewidth]{D1360_fit_expression}}
|
|
(2,1) node{\includegraphics[width=0.47\linewidth]{D1360_fit}}
|
|
|
|
(0,1.6) node {\smaller a) Target}
|
|
(1,1.6) node {\smaller b) Fit}
|
|
(2,1.6) node {\smaller c) Normalized};
|
|
\end{tikzpicture}
|
|
\vspace{0.5em}
|
|
|
|
Expression normalisation for two scans of the same individual.
|
|
The robust fitting gives a good estimate (b) of the true face surface given
|
|
the noisy measurement (a). It fills in holes and removes artifacts using
|
|
prior knowledge from the face model. The pose and expression normalized faces
|
|
(c) are used for face recognition.
|
|
\vspace{0.5em}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Funding}{name=funding,column=1,span=2,above=bottom}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\smaller
|
|
\hspace{1em}This work was supported in part by Microsoft Research through the European PhD Scholarship Programme.
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Results}{name=results,column=1,span=2,below=results neutralization,above=funding}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\begin{multicols}{2}
|
|
The method was evaluated on the GavabDB expression dataset which
|
|
contains 427 Scans, with 3 neutral scans and 4 expression scans per ID.
|
|
To test the impact of expression invariance on neutral data we used the
|
|
UND Dataset from the Face Recognition Great Vendor Test, which contains
|
|
953 neutral scans with one to eight scans per subject.
|
|
\end{multicols}\vspace{-1em}
|
|
\mbox{\hspace{0.3\linewidth}\rule{0.4\linewidth}{1pt}\hspace{0.3\linewidth}}\\
|
|
\begin{tabular}{cc}
|
|
\hspace{-0.5em}\scalebox{0.735}{\input{shrec_mncg}} &
|
|
\hspace{0.5em}\scalebox{0.735}{\input{und_mncg}}
|
|
\end{tabular}\\
|
|
% \begin{multicols}{2}
|
|
{Expression neutralization improves results on the expression dataset
|
|
without decreasing the accuracy on the neutral testset. Plotted is the
|
|
ratio of correct answers to the number of possible correct answers.
|
|
%Note the different scales for the two graphs.
|
|
%Our approach has a high accuracy on the neutral (UND) dataset.
|
|
}
|
|
% \end{multicols}\vspace{-1em}
|
|
\\\mbox{\hspace{0.3\linewidth}\rule{0.4\linewidth}{1pt}\hspace{0.3\linewidth}}\\
|
|
\begin{tabular}{cc}
|
|
\hspace{-0.5em}\scalebox{0.735}{\input{shrec_pr}} &
|
|
\hspace{0.5em}\scalebox{0.735}{\input{und_pr}}
|
|
\end{tabular}\\
|
|
% \begin{multicols}{2}
|
|
{Plotted are precision and recall for different retrieval depths. The lower
|
|
precision of the UND database is due to the fact that some queries have no
|
|
correct answers.}
|
|
% \end{multicols}\vspace{-1em}
|
|
\\\mbox{\hspace{0.3\linewidth}\rule{0.4\linewidth}{1pt}\hspace{0.3\linewidth}}\\
|
|
\begin{tabular}{cc}
|
|
\hspace{-0.5em}\scalebox{0.735}{\input{shrec_far_frr}} &
|
|
\hspace{0.5em}\scalebox{0.735}{\input{und_far_frr}}
|
|
\end{tabular}\\
|
|
% \begin{multicols}{2}
|
|
{Impostor detection is reliable, as the minimum distance to a match
|
|
is smaller than the minimum distance to a nonmatch. }
|
|
% \end{multicols}
|
|
\\
|
|
\vspace{0.5em}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Robustness}{name=robustness,column=2,row=0,above=results,span=1}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\begin{tikzpicture}[x=0.3333\linewidth,y=-0.42\linewidth]
|
|
\path [use as bounding box] (-0.5,-0.5) rectangle(2.5,1.7);
|
|
\path
|
|
(0,0) node{\includegraphics[width=0.42\linewidth]{D1160}}
|
|
(1,0) node{\includegraphics[width=0.42\linewidth]{D1425}}
|
|
(2,0) node{\includegraphics[width=0.42\linewidth]{D1205}}
|
|
|
|
(0,1) node{\includegraphics[width=0.28\linewidth]{D1160_fit_expression}}
|
|
(1,1) node{\includegraphics[width=0.28\linewidth]{D1425_fit_expression}}
|
|
(2,1) node{\includegraphics[width=0.28\linewidth]{D1205_fit_expression}}
|
|
|
|
(1,0.5) node {\smaller a) Target}
|
|
(1,1.6) node {\smaller b) Robust Reconstruction};
|
|
\end{tikzpicture}
|
|
\vspace{0.5em}
|
|
|
|
The reconstruction (b) is robust against scans (a) with artifacts, noise, and
|
|
holes.
|
|
|
|
This is achieved by a robust iteratively reweighted ICP algorithm and outlier
|
|
rejection based on angle comparisions between corresponding points.
|
|
}
|
|
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{References}{name=references,column=0,above=bottom}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\smaller
|
|
\vspace{-0.4em}
|
|
\bibliographystyle{ieee}
|
|
\renewcommand{\section}[2]{\vskip 0.05em}
|
|
\begin{thebibliography}{1}\itemsep=-0.01em
|
|
\setlength{\baselineskip}{0.4em}
|
|
\bibitem{amberg07:nonrigid}
|
|
B.~Amberg, S.~Romdhani, T. Vetter.
|
|
\newblock {O}ptimal {S}tep {N}onrigid {ICP} {A}lgorithms for {S}urface {R}egistration
|
|
\newblock In {\em Computer Vision and Pattern Recognition 2007}
|
|
\bibitem{amberg08:recognition}
|
|
B.~Amberg, R.~Knothe, T. Vetter.
|
|
\newblock Expression Invariant Face Recognition with a 3D Morphable Model
|
|
\newblock In {\em Automated Face and Gesture Recognition 2008}
|
|
\end{thebibliography}
|
|
}
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
\headerbox{Open Questions}{name=questions,column=0,span=1,below=measure,above=references}{
|
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
|
While the expression and identity space are linearly independent, there is
|
|
some expression left in the identity model. This is because a ``neutral''
|
|
face is interpreted differently by the subjects.
|
|
|
|
We investigate learning ``pure'' separated models from our the available ``impure'' data.
|
|
}
|
|
|
|
\end{poster}
|
|
|
|
\end{document}
|