KaLDeX integrates Kalman-filter linear deformable convolution and cross-attention inside UNet++ with persistent-homology loss, reporting higher accuracy than prior models on DRIVE, CHASE_DB1, STARE and OCTA-500 vessel datasets.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.IV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
KaLDeX: Kalman Filter based Linear Deformable Cross Attention for Retina Vessel Segmentation
KaLDeX integrates Kalman-filter linear deformable convolution and cross-attention inside UNet++ with persistent-homology loss, reporting higher accuracy than prior models on DRIVE, CHASE_DB1, STARE and OCTA-500 vessel datasets.