An invertible-NeRF test-time optimization approach tracks points in 2D and 3D across surgical videos, reporting roughly 50% higher average precision than prior optimization-based 2D trackers.
Metriplectic geometry for gravitational subsystems
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
In general relativity, it is difficult to localise observables such as energy, angular momentum, or centre of mass in a bounded region. The difficulty is that there is dissipation. A self-gravitating system, confined by its own gravity to a bounded region, radiates some of the charges away into the environment. At a formal level, dissipation implies that some diffeomorphisms are not Hamiltonian. In fact, there is no Hamiltonian on phase space that would move the region relative to the fields. Recently, an extension of the covariant phase space has been introduced to resolve the issue. On the extended phase space, the Komar charges are Hamiltonian. They are generators of dressed diffeomorphisms. While the construction is sound, the physical significance is unclear. We provide a critical review before developing a geometric approach that takes into account dissipation in a novel way. Our approach is based on metriplectic geometry, a framework used in the description of dissipative systems. Instead of the Poisson bracket, we introduce a Leibniz bracket - a sum of a skew-symmetric and a symmetric bracket. The symmetric term accounts for the loss of charge due to radiation. On the metriplectic space, the charges are Hamiltonian, yet they are not conserved under their own flow.
citation-role summary
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Surg-InvNeRF: Invertible NeRF for 3D tracking and reconstruction in surgical vision
An invertible-NeRF test-time optimization approach tracks points in 2D and 3D across surgical videos, reporting roughly 50% higher average precision than prior optimization-based 2D trackers.