FLOWING: Implicit Neural Flows for Structure-Preserving Morphing

Morphing is a long-standing problem in vision and computer graphics, requir-ing a time-dependent warping for feature alignment and a blending for smoothinterpolation. Recently, multilayer perceptrons (MLPs) have been explored asimplicit neural representations (INRs) for modeling such deformations, due totheir meshlessness and differentiability; however, extracting coherent and accuratemorphings from standard MLPs typically relies on costly regularizations, whichoften lead to unstable training and prevent effective feature alignment.

To overcomethese limitations, we propose FLOWING (FLOW morphING), a framework thatrecasts warping as the construction of a differential vector flow, naturally ensuringcontinuity, invertibility, and temporal coherence by encoding structural flow prop-erties directly into the network architectures. This flow-centric approach yieldsprincipled and stable transformations, enabling accurate and structure-preservingmorphing of both 2D images and 3D shapes.

Extensive experiments across arange of applications— including face and image morphing, as well as GaussianSplatting morphing— show that FLOWING achieves state-of-the-art morphingquality with faster convergence. Code and pretrained models are available in https://schardong.github.io/flowing.