================ levelX.dat ================

$r_s$	$r_t$	$currentLevel$

$sMinGeo$	$sMaxGeo$	$tMinGeo$	$tMinGeo$

$list of corresponding vertex pairs from source (left) to target (right) mesh$

-2

$D_grd$	$D_iso$

$d*_grd$	$s_i$	$t_j$

$d*_iso$	$s_i$	$t_j$

-2

for each corresponding (base) vertex pair $s_i, t_j$
$s_i$
$patch of s_i$
$t_j$
$patch of t_j$
endfor

-2

5

$5-size correspondence list extracted from the initial ~7x7 list$





=========== comments ===========

+ extra empty lines above do not appear in the real output, here for readibility

+ $r_s$ $r_t$: radii of the source and target patches of the upcoming level

+ $sMinGeo$ $sMaxGeo$ $tMinGeo$ $tMinGeo$: min and max geodesic distances on source and target meshes

+ $D_grd$ $D_iso$: ground-truth and isometric cost of computed correspondence; interpret values by considering that max geodesic is normalized to 1

+ $d*_grd$ $s_i$ $t_j$: worst match $s_i$ $t_j$ w/ ground-truth cost $d*_grd$

+ $d*_iso$ $s_i$ $t_j$: worst match $s_i$ $t_j$ w/ isometric cost $d*_iso$

+ ground-truth correspondence, if exists, assumed to be s_i to t_i where i in [0, numberOfVerticesOnMeshes]; just ignore arbitrary $D_grd$ and $d*_grd$ values if inputs do not possess this constraint

+ output of our sample run on ballerina01-ballerina02 pair is given in the folder, both textually (*.dat) and visually (allLevels.png)


================ levelX pass1.dat ================

$r_s$	$r_t$	$currentLevel$ //same as levelX.dat

$sMinGeo$	$sMaxGeo$	$tMinGeo$	$tMinGeo$ //same as levelX.dat

$list of corresponding vertex pairs from source (left) to target (right) mesh$ //not many-to-many, i.e., first step of our merging algo in sect. 4.2



================ citation ================

Y. Sahillioglu, Y. Yemez, Coarse-to-Fine Combinatorial Matching for Dense Isometric Shape Correspondence, SGP, 2011 (page # not known yet)
