ARTFEED — Contemporary Art Intelligence

4D-GSW: Kinematic-Aware Watermarking for 4D Gaussian Splatting

ai-technology · 2026-05-23

Researchers propose 4D-GSW, a kinematic-aware watermarking framework for 4D Gaussian Splatting (4DGS). 4DGS enables high-fidelity dynamic reconstruction, but protecting intellectual property of such assets is challenging. Conventional steganography neglects kinematic manifolds, causing temporal flickering and FVD collapse. 4D-GSW embeds robust copyright information while preserving spatio-temporal consistency. It introduces a Spatio-Temporal Curvature (STC) metric to identify 'Dynamic Instants' and adaptively gates watermark gradient injection to shield motion manifolds from non-physical perturbations. The framework ensures global coherence across frames. The paper is available on arXiv (2605.22342).

Key facts

  • 4D-GSW is a kinematic-aware watermarking framework for 4D Gaussian Splatting.
  • It addresses non-physical artifacts like temporal flickering and FVD collapse.
  • Introduces Spatio-Temporal Curvature (STC) metric to identify Dynamic Instants.
  • Adaptively gates watermark gradient injection to protect motion manifolds.
  • Preserves high spatio-temporal consistency while embedding copyright information.
  • Paper available on arXiv with ID 2605.22342.
  • Conventional steganography neglects kinematic manifolds.
  • 4DGS enables high-fidelity dynamic reconstruction.

Entities

Institutions

  • arXiv

Sources