Research map / 3D vision / 3D generation
Four Ways to Improve Verbo-visual Fusion for Dense 3D Visual Grounding
Ozan Unal, Christos Sakaridis, Suman Saha, Luc Van Gool
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 bottom-up attentive fusion (spherical masked attention) (ablation) | 3DVG-T with segmentation backbone · ScanRefer val | Acc@50 overall: 36.95 -> 41.08 | Improved |
| add bottom-up attentive fusion (ablation) | ConcreteNet · ScanRefer val, unique subset | Acc@50 unique: 76.64 -> 74.49 | Hurt |
| add contrastive verbo-visual loss (ablation) | ConcreteNet · ScanRefer val | Acc@50 overall: 41.08 -> 41.91 | Improved |
| add learned global camera token (ablation) | ConcreteNet · ScanRefer val | Acc@50 overall: 41.91 -> 43.84 | Improved |
| add multi-view ensembling (ablation) | ConcreteNet · ScanRefer val | Acc@50 overall: 43.84 -> 46.53 | Improved |
| replace GloVe+GRU with MPNet word encoder (ablation) | 3DVG-T with segmentation backbone · ScanRefer val | Acc@50 overall: 36.60 -> 36.95 | Improved |
| use bottom-up instead of top-down masking radius schedule (ablation) | BAF · ScanRefer val, multiple subset | Acc@50 multiple: 32.16 -> 33.43 | Improved |
| replace instance-segmentation backbone for detection output in 3DVG-T (ablation) | 3DVG-T · ScanRefer val | Acc@25 overall: Acc@50 34.47 -> 36.60; Acc@25 45.90 -> 42.29 | Mixed |
| supply ground-truth camera position as direct input instead of learning it (ablation) | ConcreteNet · ScanRefer val | Acc@50 overall: 43.84 -> 51.18 | Improved |
| extend GCT target with yaw and pitch (ablation) | GCT · ScanRefer val | Acc@50 overall: 43.84 -> 42.88 | Hurt |
Closest papers by meaning
- From Objects to Anywhere: A Holistic Benchmark for Multi-level Visual Grounding in 3D Scenes (NeurIPS 2025 · similarity 0.93)
- DenseGrounding: Improving Dense Language-Vision Semantics for Ego-centric 3D Visual Grounding (ICLR 2025 · similarity 0.91)
- CoT3DRef: Chain-of-Thoughts Data-Efficient 3D Visual Grounding (ICLR 2024 · similarity 0.90)
- Multi-Object 3D Grounding with Dynamic Modules and Language-Informed Spatial Attention (NeurIPS 2024 · similarity 0.90)
- Grounded 3D-Aware Spatial Vision-Language Modeling (CVPR 2026 · similarity 0.90)
- 3EED: Ground Everything Everywhere in 3D (NeurIPS 2025 · similarity 0.90)
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.
