Research map / 3D vision / Gaussian splatting
ARTDECO: Toward High-Fidelity On-the-Fly Reconstruction with Hierarchical Gaussian Structure and Feed-Forward Guidance
Guanghao Li, Kerui Ren, Linning Xu, Zhewen Zheng, Changjian Jiang, Xin Gao, Bo Dai, Jian Pu, Mulin Yu, Jiangmiao Pang
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 level-of-detail Gaussians (ablation) | ARTDECO · ScanNet++ rendering | PSNR: 28.13 -> 29.12 | Improved |
| add implicit structural Gaussians (ablation) | ARTDECO · ScanNet++ rendering | PSNR: 28.54 -> 29.12 | Improved |
| add global features (ablation) | ARTDECO · ScanNet++ rendering | PSNR: 28.89 -> 29.12 | Improved |
| add mapper frames (ablation) | ARTDECO · ScanNet++ rendering | PSNR: 26.38 -> 29.12 | Improved |
| add common frames (ablation) | ARTDECO · ScanNet++ rendering | PSNR: 27.20 -> 29.12 | Improved |
| add mapper frames (ablation) | ARTDECO · ScanNet++ rendering | LPIPS: 0.229 -> 0.167 | Improved |
| add loop closure (ablation) | ARTDECO · ScanNet++ tracking | ATE RMSE: 0.057 -> 0.018 | Improved |
| use MASt3R instead of pi3 as feed-forward backbone (ablation) | ARTDECO · ScanNet++ tracking | ATE RMSE: 0.374 -> 0.018 | Improved |
| use pi3 instead of VGGT in loop module (ablation) | ARTDECO · ScanNet++ tracking | ATE RMSE: 0.096 -> 0.018 | Hurt |
| track with both mapper frames and keyframes (dense) (ablation) | ARTDECO · ScanNet++ tracking | ATE RMSE: 0.018 -> 0.094 | Hurt |
| ARTDECO vs LongSplat on training time | ARTDECO · ScanNet++ | training time: 442.96 min -> 5.33 min | Improved |
| ARTDECO vs MASt3R-SLAM tracking | ARTDECO · ScanNet++ | ATE RMSE: 0.025 -> 0.018 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LongSplat — LongSplat: Robust Unposed 3D Gaussian Splatting for Casual Long Videos (ICCV 2025) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Pose-Only Solution to Visual Reconstruction and Navigation (TPAMI 2021 · similarity 0.87)
- GRF-Recon: Global Ray-Field Optimization for Long-Sequence Feed-forward Reconstruction (ECCV 2026 · similarity 0.86)
- No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images (ICLR 2025 · similarity 0.86)
- PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting for Novel View Synthesis (ICML 2025 · similarity 0.86)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.86)
- ProDyG: Progressive Dynamic Scene Reconstruction via Gaussian Splatting from Monocular Videos (NeurIPS 2025 · similarity 0.85)
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.
