Research map / 3D vision / 3D generation
Fused-Planes: Why Train a Thousand Tri-Planes When You Can Share?
Karim Kassab, Antoine Schnepf, Jean-Yves Franceschi, Laurent Caraffa, Flavian Vasile, Jeremie Mary, Andrew I. Comport, Valerie Gouet-Brunet
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 |
|---|---|---|---|
| train Fused-Planes with shared base planes (vs Tri-Planes) | Tri-Planes · ShapeNet (4 classes) | training time (min): 64.32 -> 8.96 | Improved |
| share base planes and train in latent space | Tri-Planes · ShapeNet (4 classes) | PSNR: 28.15 -> 30.47 | Improved |
| remove micro planes (ULW variant) | Fused-Planes · ShapeNet (4 classes) | per-object size (MB): 0.48 -> 0.0008 | Improved |
| use a set of base planes instead of M=1 (ablation) | Fused-Planes · ShapeNet Cars (ablation) | PSNR: 27.69 -> 28.64 | Improved |
| add macro planes (shared) in latent space (ablation) | Fused-Planes · ShapeNet Cars (ablation) | PSNR: 27.64 -> 28.64 | Improved |
| train in latent space rather than RGB (ablation) | Fused-Planes · ShapeNet Cars (ablation) | training time (min): 63.52 -> 8.92 | Improved |
| train in latent space rather than RGB (ablation) | Fused-Planes · ShapeNet Cars (ablation) | PSNR: 27.71 -> 28.64 | Improved |
| increase base planes from 50 to 75 (ablation) | Fused-Planes · ShapeNet Cars | PSNR: 30.27 -> 29.62 | Hurt |
| number of base planes M from 5 to 50 (ablation) | Fused-Planes · ShapeNet Cars | PSNR: 29.89 -> 30.27 | Improved |
| use low-budget randomly initialized VAE (ablation) | Fused-Planes · ShapeNet Cars | PSNR: 30.27 -> 29.22 | Hurt |
| train on 4 classes jointly instead of 1 | Fused-Planes · Speakers | PSNR: 29.99 -> 29.72 | Hurt |
| Fused-Planes vs K-Planes quality | K-Planes · ShapeNet (4 classes) | PSNR: 30.88 -> 30.47 | Hurt |
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.
- CodeNeRF — CodeNeRF: Disentangled Neural Radiance Fields for Object Categories (ICCV 2021) · better on 1 of 1 shared result
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
- Towards In-the-wild 3D Plane Reconstruction from a Single Image (CVPR 2025 · similarity 0.87)
- AirPlanes: Accurate Plane Estimation via 3D-Consistent Embeddings (CVPR 2024 · similarity 0.86)
- AlphaTablets: A Generic Plane Representation for 3D Planar Reconstruction from Monocular Videos (NeurIPS 2024 · similarity 0.86)
- Neural Processing of Tri-Plane Hybrid Neural Fields (ICLR 2024 · similarity 0.85)
- HyPlaneHead: Rethinking Tri-plane-like Representations in Full-Head Image Synthesis (NeurIPS 2025 · similarity 0.85)
- Large-Vocabulary 3D Diffusion Model with Transformer (ICLR 2024 · 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.
