1 paper · 1 filter
Mathias Lowes, Kristine Aavild Sørensen, Maxime Sermesant +1
Implicit Neural Representations (INRs) have recently gained popularity for their ability to model functions using simple and lightweight networks. These methods have also proven ef…