Learning Dynamic Scene Reconstruction with Sinusoidal Geometric Priors
This work addresses challenges in dynamic scene reconstruction for computer vision applications, but it appears incremental as it builds on existing methods with a novel loss function.
The paper tackled the problem of dynamic 3D scene reconstruction by proposing SirenPose, a loss function that combines sinusoidal activations with geometric priors, resulting in significant improvements in spatiotemporal consistency metrics for handling rapid motion and complex scenes.
We propose SirenPose, a novel loss function that combines the periodic activation properties of sinusoidal representation networks with geometric priors derived from keypoint structures to improve the accuracy of dynamic 3D scene reconstruction. Existing approaches often struggle to maintain motion modeling accuracy and spatiotemporal consistency in fast moving and multi target scenes. By introducing physics inspired constraint mechanisms, SirenPose enforces coherent keypoint predictions across both spatial and temporal dimensions. We further expand the training dataset to 600,000 annotated instances to support robust learning. Experimental results demonstrate that models trained with SirenPose achieve significant improvements in spatiotemporal consistency metrics compared to prior methods, showing superior performance in handling rapid motion and complex scene changes.