Conceptos Básicos
Holistic urban scene understanding through 3D Gaussian splatting for real-time rendering and accurate semantic reconstruction.
Resumen
Introduction to the challenging problem of holistic urban scene understanding based on RGB images.
Novel pipeline utilizing 3D Gaussian Splatting for joint optimization of geometry, appearance, semantics, and motion.
Ability to render new viewpoints in real-time with high accuracy in 2D and 3D semantic information.
Experimental results demonstrating effectiveness on various benchmarks like KITTI, KITTI-360, and Virtual KITTI 2.
Contributions include addressing dynamic 3D urban scene understanding, decomposing static and dynamic objects without ground truth bounding boxes, achieving state-of-the-art performance in novel view appearance and semantic synthesis.
Method overview illustrating decomposition of scenes into static regions and rigidly moving dynamic objects using 3D Gaussians.
Estadísticas
我々の方法は、新しい視点をリアルタイムでレンダリングする能力を提供します。
実験結果は、KITTI、KITTI-360、およびVirtual KITTI 2などのさまざまなベンチマークでの効果を示しています。