C3G is a novel feed-forward framework that efficiently reconstructs and understands 3D scenes from sparse multi-view images by estimating a compact set of 2,048 3D Gaussians. This approach significantly reduces memory overhead and improves multi-view feature aggregation compared to dense, per-pixel methods, leading to superior performance in novel view synthesis and 3D open-vocabulary segmentation.

This paper introduces CrashSplat, an automatic car damage detection pipeline that uplifts 2D damage masks to a 3D Gaussian Splatting reconstruction. It proposes a novel learning-free, single-view 3D segmentation method, offering a more geometrically accurate representation of vehicle damage compared to traditional 2D analyses.