This is a preprint describing a novel technical method for radar rendering with no peer review, clinical validation, or comparison against ground truth radar measurements—it demonstrates algorithmic feasibility on a small dataset but does not establish clinical or operational utility.
Reported
Mean Pearson correlation (RA imag…0.587
Performance gain vs RadarSplat ba…1.7× to 5.2×
Training time per sceneapproximately 3 minutes