GLUI and Visual Studio .Net 2003

When compiling a C++ program depending on glui32.lib, I got error like this:

error LNK2005: _free already defined in LIBCD.lib(dbgheap.obj)

The problem can be solved by exclude LIBCD dependency in the project settings.

Needs clarification

Proof of Linear Transformation of a multivariate normal distribution.

see Kingman and Taylor 1966, p.372

Should there be an erreta of missing a transpose sign?

See the red circle in this picture.

Worth Reading

Multivariable Calculus
George Cain & James Herod
http://www.math.gatech.edu/~cain/notes/calculus.html

comments: good to read for clarification of Jacobian matrix, total derivative and variable substitution thereoms.
Measure Theory surely works, but this online book is more concise and readable.

Obviously, the books I read before is a little outdated. How could they totally neglect multivariate calculus?

Human Eva, Bad motion capture data?

It seems that some of the motion data sequence, 'ThrowCatch' in HumanEva I is bad.

Worth Reading

Y. Lipman. Differential coordinates for interactive mesh editing.
O. Sorkin. Differential representations for mesh processing.
Edilson de Aguiar. Rapid Animation of Laser-scanned Humans.

comment: an interesting representation of mesh and morphology of mesh.

Worth Reading

Fast Human Pose Estimation using Appearance and Motion via Multi-Dimensional Boosting Regression.

comment: Another idea of me that has been worked.

My thoughts are:

The point here is using the 2D appearance and motion hue to infer the 3D state space.

Actually, because the 3D motion is a high-dimension state space and actually lies on some lower space manifold, and there are generalized regression or interpolation methods that can be used for mapping between the space and manifold. So we can just use the lower dimensional information to infer its 3D state.

This might not be an accurate method, but provides a way to speed up the motion tracking process.

The work is incomplete, and could be expanded on a lot of aspects.

I have many ideas

But most of them have been worked on.
And lots of the rest die in my lab.

frustrating....

Read Rob Hess' SIFT data into MATLAB

%% read sift features from Rob Hess output
%% compatible with Rob's version version 1.1.1-20070913 (2006)
%% see imgfeatures.c in Rob's code for more details
%%
%% by smartnose@gmail.com, Jan. 2nd 2008
%%
%% INPUTS:
%% fn - full path string to the file to be read
%%
%% OUTPUTS:
%% rs - result structure
%% n: number of features
%% d: dimension of the feature vector (128 by default)
%% sv: scale and feature position vectors, each row indicate
%% y,x,scale, orientation of the feature.
%% fv: feature vectors, each row is a vector


function [rs]=import_rob_sift(fn)
if(~exist(fn,'file'))
error('file does not exist');
end

M=dlmread(fn);%M will be a N*20 matrix, with blank cells filled by 0

rs.n=M(1,1);
rs.d=M(1,2);

SV=zeros(rs.n,4);
FV=zeros(rs.n,rs.d);

ln=ceil(rs.d/20);

for ii=1:rs.n
SV(ii,:)=M((ii-1)*8+2,1:4);
%FV(ii,:)=
FT=[];
for jj=2:ln+1
FT=cat(2,FT,M((ii-1)*8+jj+1,:));
end
FV(ii,:)=FT(1,1:rs.d);
end

rs.sv=SV;
rs.fv=FV;
end