Główne pojęcia
Periodic Vibration Gaussian (PVG) enhances dynamic urban scene reconstruction and rendering efficiency.
Streszczenie
The article introduces PVG, a model for dynamic urban scene representation. It addresses challenges in modeling dynamic scenes by introducing periodic vibration-based temporal dynamics. PVG surpasses state-of-the-art alternatives in reconstruction and novel view synthesis without relying on manual annotations or optical flow estimation. The model achieves significant efficiency benefits in training and rendering processes.
- Introduction to the challenges of modeling dynamic urban scenes.
- Description of the Periodic Vibration Gaussian (PVG) model.
- Explanation of the key features of PVG, including dynamics introduction and periodic vibration.
- Details on position-aware point adaptive control strategy and temporal smoothing mechanism.
- Results from experiments on Waymo Open Dataset and KITTI benchmarks showcasing PVG's superiority over competitors.
- Ablation study results highlighting the impact of different components on novel view synthesis.
Statystyki
PVG surpasses state-of-the-art alternatives in both reconstruction and novel view synthesis for both dynamic and static scenes.
PVG exhibits 900-fold acceleration in rendering over the best alternative.