Research map / 3D vision / Gaussian splatting
GaussianPile: A Unified Sparse Gaussian Splatting Framework for Slice-based Volumetric Reconstruction
Di Kong, Yikai Wang, Wenjie Guo, Yifan Bu, Boya Zhang, Yuexin Duan, Xiawei Yue, Wenbiao Du, Yiman Zhong, Yuwen Chen, Cheng Ma
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 |
|---|---|---|---|
| use the focus-aware finite-thickness rendering model instead of slice-adapted 3DGS (ablation) | GaussianPile · ABUS (3D volume) | PSNR: 28.49 -> 33.22 | Improved |
| use the focus-aware model instead of slice-adapted 3DGS (ablation) | GaussianPile · hiPSC (3D volume) | PSNR: 15.20 -> 33.73 | Improved |
| use the focus-aware model instead of slice-adapted 3DGS (ablation) | GaussianPile · ABUS (2D slices) | PSNR: 27.41 -> 33.07 | Improved |
| add Morton-order quantization and LZMA entropy coding (ablation) | GaussianPile · ABUS (memory) | memory size (MB): 49.8 -> 6.2 | Improved |
| set axial scale equal to the scanning step instead of one tenth of it (ablation) | GaussianPile · Tribolium high | PSNR: 34.25 -> 36.67 | Improved |
| use axial scale equal to scanning step instead of 10 times it (ablation) | GaussianPile · Tribolium high | PSNR: 23.05 -> 36.67 | Improved |
| increase initial Gaussians from 1000k to 2000k (ablation) | GaussianPile · Tribolium high | PSNR: 36.67 -> 36.82 | No clear effect |
| decrease initial Gaussians from 1000k to 50k (ablation) | GaussianPile · Tribolium high | PSNR: 36.67 -> 34.12 | Hurt |
| use grid initialization instead of random initialization (ablation) | GaussianPile · Tribolium high | PSNR: 36.67 -> 36.30 | Hurt |
| train 30k iterations instead of 10k (ablation) | GaussianPile · ABUS (2D slices) | PSNR: 32.25 -> 33.07 | Improved |
| drop spherical harmonics coefficients from Gaussian primitives | GaussianPile · parameter count per primitive | parameters per primitive: approximately 40% fewer | Improved |
Closest papers by meaning
- R$^2$-Gaussian: Rectifying Radiative Gaussian Splatting for Tomographic Reconstruction (NeurIPS 2024 · similarity 0.91)
- Learning Efficient Fuse-and-Refine for Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.90)
- Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis (CVPR 2024 · similarity 0.89)
- Volumetrically Consistent 3D Gaussian Rasterization (CVPR 2025 · similarity 0.89)
- CompactSplat: Spatially Adaptive Gaussian Distribution for Feedforward Scene Reconstruction (NeurIPS 2026 · similarity 0.89)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.88)
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.
