StudioRecon: 4D Human-Scene Reconstruction from Low-Overlap Captures

StudioRecon targets 4D human-scene reconstruction when only a few cameras with low overlap are available. The pipeline decouples backgrounds from humans, synthesizes hundreds of camera-controlled novel views with a video diffusion model to densify background supervision, and initializes deformable Gaussian humans using cross-view identity association plus triangulated multi-view keypoint fitting. A recursive enhancement module with motion-adaptive consistency injection then harmonizes the composition and reduces residual artifacts. The authors report state-of-the-art novel-view synthesis on four real-world datasets and demonstrate novel trajectory rendering and human replacement.