Research map / 3D vision / Gaussian splatting
B$^3$-Seg: Camera-Free, Training-Free 3DGS Segmentation via Analytic EIG and Beta-Bernoulli Bayesian Updates
Hiromichi Kamata, Samuel Arthur Munro, Fuminori Homma
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add CLIP re-ranking of candidate masks (ablation) | B3-Seg · LERF-Mask | mIoU: 74.9 -> 81.5 (+6.6) | Improved |
| add SAM2 mask-input prior on top of CLIP re-ranking (ablation) | B3-Seg · LERF-Mask | mIoU: 81.5 -> 84.5 (+3.0) | Improved |
| add SAM2 mask-input prior on top of CLIP re-ranking (ablation) | B3-Seg · LERF-Mask | mBIoU: 76.6 -> 81.0 | Improved |
| increase candidate views N_cand from 5 to 20 (ablation) | B3-Seg · LERF-Mask | mIoU: 76.2 -> 84.5 | Improved |
| increase candidate views N_cand from 20 to 30 (ablation) | B3-Seg · LERF-Mask | mIoU: 84.5 -> 84.6; time 12.1 s -> 14.4 s | No clear effect |
| increase iterations T from 5 to 20 (ablation) | B3-Seg · LERF-Mask | mIoU: 79.8 -> 84.5; time 3.5 s -> 12.1 s | Improved |
| increase iterations T from 20 to 30 (ablation) | B3-Seg · LERF-Mask | mIoU: 84.5 -> 84.8; time 12.1 s -> 19.4 s | No clear effect |
| shift initial object center by 50% of r_obj (ablation) | B3-Seg · LERF-Mask | mIoU: 84.5 -> 82.9 (-1.6) | Hurt |
| shift initial object center by 100% of r_obj (ablation) | B3-Seg · LERF-Mask | mIoU: 84.5 -> 80.7 (-3.8) | Hurt |
| select views by analytic EIG instead of uniform-sphere random views (FlashSplat pipeline) | B3-Seg · LERF-Mask | mIoU: 69.6 -> 84.5 | Improved |
| select views by analytic EIG instead of random reconstruction-camera views | B3-Seg · LERF-Mask | mIoU: 76.5 -> 84.5 | Improved |
| select views by analytic EIG instead of random reconstruction-camera views | B3-Seg · 3D-OVS (Sofa) | mIoU: 85.7 -> 94.1 | Improved |
Closest papers by meaning
- Bootstraping Clustering of Gaussians for View-consistent 3D Scene Understanding (AAAI 2025 · similarity 0.88)
- Probabilistic Interactive 3D Segmentation with Hierarchical Neural Processes (ICML 2025 · similarity 0.87)
- Efficient Decoupled Feature 3D Gaussian Splatting via Hierarchical Compression (CVPR 2025 · similarity 0.87)
- LP-3DGS: Learning to Prune 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- 3DGS-Enhancer: Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors (NeurIPS 2024 · similarity 0.87)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
