DENSER: Depth-Guided Ensemble with Staged EFA-GS Reconstruction for Soccer Novel View Synthesis
Researchers have developed DENSER, a novel approach for synthesizing new views of soccer matches using depth guidance and staged reconstruction. This method incorporates camera-height-based loss weighting, monocular depth supervision from Depth-Anything-V2 to improve geometry in textureless areas, and a three-model ensemble for enhanced reconstruction. DENSER achieved strong performance on five test scenes, with reported metrics including a mean PSNR of 29.89 dB, SSIM of 0.791, and LPIPS of 0.366. AI
IMPACT Introduces a novel method for synthesizing new views in sports broadcasts, potentially improving replay quality and analysis tools.