TABLE I:

Summary of notation used in the article.

I, $\mathcal{I}=\{I_1,\dots ,I_n\}$	Test and training images
S, $\mathcal{S}=\{S_1,\dots ,S_n\}$	Test and training segmentations
Sl	Segmentation probability map for label l
$\widehat{S}$	Inferred segmentation
F, $\mathcal{F}$	Keypoint in test and training image
Fx, ${\mathcal{F}}^x$	Location of test and training keypoint
Fσ, ${\mathcal{F}}^{\sigma }$	Scale of test and training keypoint
FD, ${\mathcal{F}}^D$	Descriptor of test and training keypoint
FI, ${\mathcal{F}}_{I_i}$	All keypoints from test and training image i
L, $\mathcal{L}$	Label of test and training keypoint
$\widehat{L}$	Inferred label for test keypoint
m	Match between a test and training keypoint
m(F)	Corresponding training keypoint matched to F
$\mathcal{M}\left(F\right)$	All matches for test keypoint F
${\mathcal{M}}_i$	All matches for training image Ii
${\mathcal{I}}_m$, ${\mathcal{S}}_m$	Training image and segmentation of match m