Reconstructing a dynamic scene from image inputs is a fundamental computer vi-
sion task with many downstream applications. Despite recent advancements, ex-
isting approaches still struggle to achieve high-quality reconstructions from unseen
viewpoints and timestamps. This work introduces the ReMatching framework, de-
signed to improve reconstruction quality by incorporating deformation priors into
dynamic reconstruction models. Our approach advocates for velocity-field-based
priors, for which we suggest a matching procedure that can seamlessly supplement
existing dynamic reconstruction pipelines. The framework is highly adaptable and
canbeappliedtovariousdynamicrepresentations. Moreover, itsupportsintegrating
multiple types of model priors and enables combining simpler ones to create more
complex classes. Our evaluations on popular benchmarks involving both synthetic
andreal-worlddynamicscenesdemonstratethataugmentingcurrentstate-of-the-art
methods with our approach leads to a clear improvement in reconstruction accuracy.
|
