Research map / Point clouds / Segmentation and detection
Spatially Invariant Unsupervised 3D Object-Centric Learning and Scene Decomposition
Tianyu Wang, Miaomiao Liu, Kee Siong Ng
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 multi-layer PointGNN for z_pres generation (ablation) | SPAIR3D · UOR | ARI: 0.841 -> 0.915 | Improved |
| add multi-layer PointGNN for z_pres generation (ablation) | SPAIR3D · UOR | SC: 0.610 -> 0.832 | Improved |
| add multi-layer PointGNN for z_pres generation (ablation) | SPAIR3D · UOR | mSC: 0.627 -> 0.836 | Improved |
| change voxel size to 0.75l (ablation) | SPAIR3D · UOR | ARI: 0.915 -> 0.932 | No clear effect |
| change voxel size to 1.25l (ablation) | SPAIR3D · UOR | ARI: 0.915 -> 0.922 | No clear effect |
| evaluate on 6-12 object scenes unseen in training | SPAIR3D · UOR | ARI: 0.915 -> 0.912 | No clear effect |
| evaluate on 6-12 object scenes unseen in training | SPAIR3D · UOT | ARI: 0.901 -> 0.892 | No clear effect |
| evaluate on 16-object Object Matrix scenes (trained on 2-5 objects) | SPAIR3D · UOR | ARI: 0.915 -> 0.872 | Hurt |
| evaluate on 16-object Object Matrix scenes (trained on 2-5 objects) | SPAIR3D · UOT | ARI: 0.901 -> 0.879 | No clear effect |
| use supervised PointGroup instead of unsupervised SPAIR3D | SPAIR3D · UOR | ARI: 0.915 -> 0.976 | Improved |
| tune Mean-shift bandwidth to 0.06 (chair size) instead of 0.15 (ablation) | Mean-shift · S3DIS Chair | mIoU: 0.33 -> 0.75 | Mixed |
| tune Mean-shift bandwidth to 0.15 (table size) instead of 0.06 (ablation) | Mean-shift · S3DIS Table | mIoU: 0.34 -> 0.46 | Improved |
Closest papers by meaning
- PointGLR: Unsupervised Structural Representation Learning of 3D Point Clouds (TPAMI 2022 · similarity 0.89)
- Global-Local Bidirectional Reasoning for Unsupervised Representation Learning of 3D Point Clouds (CVPR 2020 · similarity 0.88)
- RfD-Net: Point Scene Understanding by Semantic Instance Reconstruction (CVPR 2021 · similarity 0.88)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.87)
- UniPre3D: Unified Pre-training of 3D Point Cloud Models with Cross-Modal Gaussian Splatting (CVPR 2025 · similarity 0.87)
- DOCTR: Disentangled Object-Centric Transformer for Point Scene Understanding (AAAI 2024 · similarity 0.86)
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.
