Research map / 3D vision / 3D generation
PrITTI: Primitive-based Generation of Controllable and Editable 3D Semantic Urban Scenes
Christina Ourania Tze, Daniel Dauner, Yiyi Liao, Dzmitry Tsishkou, Andreas Geiger
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 latent split in the LVAE (ablation) | PrITTI LVAE · KITTI-360 LVAE ablation | Raster MSE (x10^-2): 3.55 -> 0.75 | Improved |
| add latent split in the LVAE (ablation) | PrITTI LVAE · KITTI-360 LVAE ablation | AP3D: 53.78 -> 62.12 | Improved |
| add latent split in the LVAE (ablation) | PrITTI LVAE · KITTI-360 LVAE ablation | AP3D@50: 39.09 -> 46.96 | Improved |
| joint training with ground raster branch (ablation) | PrITTI LVAE · KITTI-360 LVAE ablation | AP3D: 60.28 -> 62.12 | Improved |
| joint training with object branch (ablation) | PrITTI LVAE · KITTI-360 LVAE ablation | Raster MSE (x10^-2): 0.59 -> 0.75 | Hurt |
| increase SemCity decoder query points from 400K to 1M | SemCity · KITTI-360 voxel reconstruction | mIoU: 70.52 -> 93.81 | Improved |
| reduce voxel resolution then upsample | SemCity · KITTI-360 voxel reconstruction | mIoU: 70.52 -> 50.52 | Hurt |
| primitive representation instead of voxels | PrITTI · KITTI-360 voxel reconstruction | Size (MB): 8.00 -> 2.52 | Improved |
| scale DiT-B to DiT-L | PrITTI · KITTI-360 3D scene generation | FID: 73.952 -> 71.028 | Improved |
| scale DiT-L to DiT-XL | PrITTI · KITTI-360 3D scene generation | FID: 71.028 -> 71.735 | No clear effect |
| PrITTI instead of XCube Level 2 | PrITTI DiT-B · KITTI-360 3D scene generation | FID: 94.822 -> 73.952 | Improved |
| Cholesky encoding instead of quaternions (ablation) | PrITTI object encoding · synthetic vehicle primitives | IoU3D | Improved |
Closest papers by meaning
- SemCity: Semantic Scene Generation with Triplane Diffusion (CVPR 2024 · similarity 0.89)
- CityDreamer: Compositional Generative Model of Unbounded 3D Cities (CVPR 2024 · similarity 0.88)
- SceneFactor: Factored Latent 3D Diffusion for Controllable 3D Scene Generation (CVPR 2025 · similarity 0.87)
- A 3D Scene is Worth 1K Tokens: 3D-Grounded Representation for Scene Generation at Scale (NeurIPS 2026 · similarity 0.87)
- CymbaDiff: Structured Spatial Diffusion for Sketch-based 3D Semantic Urban Scene Generation (NeurIPS 2025 · similarity 0.87)
- DORSal: Diffusion for Object-centric Representations of Scenes $\textit{et al.}$ (ICLR 2024 · similarity 0.87)
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.
