Research map / 3D vision / Face and shape
Iterative Superquadric Recomposition of 3D Objects from Multiple Views
Stephan Alaniz, Massimiliano Mancini, Zeynep Akata
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 iterative superquadric addition instead of random simultaneous init (ablation) | ISCO · ShapeNet (16 views) | mean volumetric IoU: 0.578 -> 0.656 | Improved |
| use iterative addition (ablation) | ISCO · ShapeNet (4 views) | mean volumetric IoU: 0.493 -> 0.576 | Improved |
| use iterative addition (ablation) | ISCO · ShapeNet (8 views) | mean volumetric IoU: 0.557 -> 0.637 | Improved |
| increase number of views from 4 to 16 (ablation) | ISCO · ShapeNet | mean volumetric IoU: 0.576 -> 0.637 -> 0.656 | Improved |
| increase the number of superquadrics (ablation) | ISCO · ShapeNet | mean IoU: gradual improvement with diminishing returns | Improved |
| set lambda to 0.6 instead of 0.4 (ablation) | ISCO · ShapeNet (16 views) | mean volumetric IoU: 0.629 -> 0.650 | Improved |
| set lambda to 0.6 instead of 0.8 (ablation) | ISCO · ShapeNet (16 views) | mean volumetric IoU: 0.613 -> 0.650 | Improved |
| use multi-view ISCO instead of point-cloud NBP | ISCO · ShapeNet | mean volumetric IoU: 0.588 -> 0.656 | Improved |
| use ISCO instead of EMS | ISCO · ShapeNet | mean volumetric IoU: 0.330 -> 0.656 | Improved |
| use ISCO instead of implicit NeuS | ISCO · ShapeNet | mean volumetric IoU: 0.701 -> 0.656 | Hurt |
| use ISCO instead of EMS/NBP with consistent assignment | ISCO · ShapeNet-part table | per-part mIoU: 32.3 -> 70.7 (EMS), 32.9 -> 70.7 (NBP) | Improved |
| train HSQ with 16 views instead of single view | HSQ · ShapeNet | mIoU: improves slightly by 0.029; ISCO 0.656 still higher | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- NeuS — NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction (NeurIPS 2021) · better on 0 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
- SuperDec: 3D Scene Decomposition with Superquadrics Primitives (ICCV 2025 · similarity 0.89)
- SuperPrimitive: Scene Reconstruction at a Primitive Level (CVPR 2024 · similarity 0.86)
- QuadricFormer: Scene as Superquadrics for 3D Semantic Occupancy Prediction (NeurIPS 2025 · similarity 0.85)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.85)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.85)
- MESC-3D:Mining Effective Semantic Cues for 3D Reconstruction from a Single Image (CVPR 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.
