ParaScale: Scale-Calibrated Camera-Motion Transfer via a Gauge-Invariant Parallax Number
Researchers have developed ParaScale, a novel module designed to accurately transfer camera motion between videos, even when the scenes are at vastly different scales. The system utilizes a gauge-invariant 'Parallax Number' to quantify camera movement, ensuring that the transferred motion is neither too subtle nor excessively exaggerated. ParaScale functions as a plug-and-play component that can be integrated into existing pose-conditioned generators without requiring retraining, and it introduces a new metric, the Parallax Consistency Error (PCE), to better detect scale mismatches. AI
IMPACT Introduces a novel method for scale-invariant camera motion transfer in generative models.