arxivcs.CV2026-07-08
Learning to Unify Deformable Shape and Texture Representations for Cardiac Video Classification
Tonmoy Hossain, Miaomiao Zhang
Deformable shape representations have proven to be robust complements to texture features in cardiac image classification, offering geometric priors that are invariant to imaging artifacts and intensity variations. However, existing deep networks perform simple concatenation to c…