Research map / 3D vision / 3D generation
ClusteringSDF: Self-Organized Neural Implicit Surfaces for 3D Decomposition
Tianhao Wu, Chuanxia Zheng, Qianyi Wu, Tat-Jen Cham
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 clustering loss to ObjectSDF++ even with ground-truth instance labels (ablation) | ObjectSDF++ · Replica | edge accuracy: 28.9 -> 58.9 | Improved |
| add clustering loss to ObjectSDF++ with GT labels (ablation) | ObjectSDF++ · Replica | PQ scene: 64.2 -> 69.1 | Improved |
| replace ObjectSDF++ opacity supervision with clustering method on Mask2Former labels (ablation) | ObjectSDF++ · Replica | PQ scene: 43.8 -> 59.2 | Improved |
| add regularization loss (variance of softmax values) (ablation) | ClusteringSDF · Replica | PQ scene: 52.6 -> 59.2 | Improved |
| add one-hot loss (ablation) | ClusteringSDF · Replica | PQ scene: 56.3 -> 59.2 | Improved |
| add foreground-background loss (ablation) | ClusteringSDF · Replica | PQ scene: 57.7 -> 59.2 | Improved |
| remove differentiation loss (ablation) | ClusteringSDF · Replica | segmentation | Hurt |
| add cross-view loss (ablation) | ClusteringSDF · ScanNet/Replica instances never co-visible | instance label assignment | Improved |
| use clustering loss instead of pixel-pair contrastive loss (Contrastive Lift) | ClusteringSDF · ScanNet/Replica | training time: >30h -> <8h | Improved |
| use ClusteringSDF instead of Contrastive Lift | ScanNet · ScanNet | PQ scene: 62.3 -> 61.0 | Hurt |
Closest papers by meaning
- ObjectSDF++: Improved Object-Compositional Neural Implicit Surfaces (ICCV 2023 · similarity 0.89)
- Local Deep Implicit Functions for 3D Shape (CVPR 2020 · similarity 0.89)
- Learning to Infer Implicit Surfaces without 3D Supervision (NeurIPS 2019 · similarity 0.88)
- LP-DIF: Learning Local Pattern-Specific Deep Implicit Function for 3D Objects and Scenes (CVPR 2023 · similarity 0.88)
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 2024 · similarity 0.87)
- Total-Decom: Decomposed 3D Scene Reconstruction with Minimal Interaction (CVPR 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.
