Research map / 3D vision / 3D generation
Multi-view Invariance Learning for 3D Scene Graph Pre-training via Collaborative Cross-Modal Regularization
Yucheng Huang, Luping Ji, Ruijie Xiao, Jiayuan Sun
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 cross-modal regularization (CM) to Barlow Twins pre-training (ablation) | SGFN PointNet · 3DSSG | predicate mA@1: 48.63 -> 53.69 | Improved |
| add DEC clustering loss on top of BT+CM (ablation) | SGFN PointNet · 3DSSG | predicate mA@1: 53.69 -> 56.10 | Improved |
| add knowledge-filtering gate adapter (ablation) | SGFN PointNet · 3DSSG | predicate mA@1: 45.43 -> 56.10 | Improved |
| replace gated fusion with element-wise addition (ablation) | SGFN PointNet · 3DSSG | predicate mA@1: 56.10 -> 52.34 | Hurt |
| Barlow Twins invariance only vs no pre-training (ablation) | SGFN PointNet · 3DSSG | object A@1: 53.67 -> 55.35 | Improved |
| pre-training batch size 128 vs 64/256 (ablation) | SGFN PointNet w/ PR · 3DSSG | triplet mA@50: 60.48/65.16 -> 65.82 | Improved |
| double pre-training data (4672 -> 9208 segments) (ablation) | SGFN PointNet w/ PR · 3DSSG | triplet mA@50: 62.87 -> 65.82 | Improved |
| our pre-training vs DepthContrast | VL-SAT PointNet++ · 3DSSG | predicate mA@1: 52.68 -> 55.32 | Improved |
| DepthContrast pre-training vs none | VL-SAT PointNet++ · 3DSSG | predicate mA@1: 52.69 -> 52.68 | No clear effect |
| add pre-training and adapter | VL-SAT PointTransformer · 3DSSG | predicate mA@1: 52.86 -> 58.43 | Improved |
| add pre-training to VL-SAT | VL-SAT PointNet · 3DSSG horizontal predicates | predicate A@1 | Mixed |
Closest papers by meaning
- Inst3D-LMM: Instance-Aware 3D Scene Understanding with Multi-modal Instruction Tuning (CVPR 2025 · similarity 0.89)
- Building a Strong Pre-Training Baseline for Universal 3D Large-Scale Perception (CVPR 2024 · similarity 0.88)
- CLIP-Driven Open-Vocabulary 3D Scene Graph Generation via Cross-Modality Contrastive Learning (CVPR 2024 · similarity 0.88)
- Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · similarity 0.88)
- Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models (NeurIPS 2026 · similarity 0.88)
- CrossOver: 3D Scene Cross-Modal Alignment (CVPR 2025 · 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.
