3D Gaussian Splatting for Fine-grained Urban Scene Reconstruction via Semantic-guided Density Control and Loss Design

Key points

• However, it still suffers from coarse density control and semantically blind loss functions when reconstructing the fine-grained structures in complex urban scenes.1
• Thus, this study proposes SG-GS, a semantic-guided 3DGS method for fine-grained structure reconstruction.1
• On the one hand, a semantic-constrained adaptive density control mechanism is constructed, which dynamically allocates gradient thresholds to different semantic regions of the scene and introduces an additional boundary-enhanced densification strategy, so that new Gaussians are directed to structurally critical regions.1
• The method can serve the fine-grained 3D reconstruction of urban scenes.1

Sources (1)

• [1]3D Gaussian Splatting for Fine-grained Urban Scene Reconstruction via Semantic-guided Density Control and Loss Design
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 12, 07:17 AM

  3D Gaussian Splatting (3DGS) has been becoming a promising 3D reconstruction technique because of its advantage of high-fidelity rendering effect with high efficiency.

  However, it still suffers from coarse density control and semantically blind loss functions when reconstructing the fine-grained structures in complex urban scenes.

Extractive summary: sentences quoted from the sources.