Research map / 3D vision / 3D generation
Sparse auto-regressive modeling for scene generation from multi-view images
Thomas Lucas, Maxime Pietrantoni, Wonjune Cho, Bardienus Duisterhof, Philippe Weinzaepfel, Vincent Leroy, Jerome Revaud
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 |
|---|---|---|---|
| generative sparse voxel latent AR model vs best baseline | SPAR3S · 3DFront 2-view | FID: 110 -> 59 (vs DepthSplat) | Improved |
| generative sparse voxel latent AR model vs LatentSplat | SPAR3S · RealEstate10k 2-view | PSNR: 16.09 -> 16.72 | Improved |
| add bincount-based guided attention (ablation) | SPAR3S encoder · 3DFront | training speed and PSNR | Improved |
| increase number of Gaussians per voxel (ablation) | SPAR3S decoder · 3DFront | PSNR | Improved |
| inject about 25% false-positive voxels at training upsampling (ablation) | SPAR3S encoder · 3DFront | PSNR | Improved |
| KL coefficient above 0.1 or latent dim below 32 (ablation) | SPAR3S encoder · 3DFront | PSNR | Hurt |
| replace diffusion head with linear head (ablation) | SPAR3S AR model · 3DFront | novel view quality | Hurt |
| add 3D relative positional encoding (ablation) | SPAR3S AR model · 3DFront | novel view quality | Improved |
| use BFS region-growing ordering instead of random order (ablation) | SPAR3S inference · 3DFront | novel view quality | Improved |
| add occupancy refinement (coarse-to-fine) (ablation) | SPAR3S · 3DFront | novel view quality | Improved |
| increase number of conditioning views 1 to 12 | SPAR3S · 3DFront | PSNR / LPIPS: PSNR 11.73 -> 19.62 at threshold 0.1 (1 -> 12 views) | Improved |
| use ground truth occupancy (ablation) | SPAR3S · 3DFront | novel view quality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 5 of 5 on most shared results.
- DepthSplat — DepthSplat: Connecting Gaussian Splatting and Depth (CVPR 2025) · better on 8 of 8 shared results
- MVSplat360 — MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024) · better on 8 of 8 shared results
- LatentSplat — latentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction (ECCV 2024) · better on 8 of 8 shared results
- PixelSplat — pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction (CVPR 2024) · better on 5 of 5 shared results
- DiffusioNeRF — DiffusioNeRF: Regularizing Neural Radiance Fields With Denoising Diffusion Models (CVPR 2023) · better on 3 of 3 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
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.90)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.89)
- MVSplat360: Feed-Forward 360 Scene Synthesis from Sparse Views (NeurIPS 2024 · similarity 0.89)
- UMAMI: Unifying Masked Autoregressive Models and Deterministic Rendering for View Synthesis (NeurIPS 2025 · similarity 0.89)
- Novel View Synthesis from A Few Glimpses via Test-Time Natural Video Completion (NeurIPS 2025 · similarity 0.89)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.89)
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.
