Research map / 3D vision / Depth and stereo
Abstract the Layout, Focus the Detail: A Dual-Granularity Representation Framework for Zero-Shot 3D Visual Grounding
Zeyuan Lin, Hanxuan Li, Chen He, Ruiping Wang, Zhaoxiang Liu, Xilin CHEN
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 query filtering and semantic spatial layout over raw BEV (ablation) | dual-granularity framework · Nr3D subset | accuracy: 48.8 -> 60.8 | Improved |
| add visual detail patches to filtering + SSL (ablation) | dual-granularity framework · Nr3D subset | accuracy: 60.8 -> 66.0 | Improved |
| add semantic spatial layout to filtering + VDP (ablation) | dual-granularity framework · Nr3D subset | accuracy: 56.0 -> 66.0 | Improved |
| filtering + VDP only vs raw BEV (ablation) | dual-granularity framework · Nr3D subset | accuracy: 48.8 -> 56.0 | Improved |
| use dual-granularity layout vs best zero-shot baseline | dual-granularity framework · ScanRefer Multiple | Acc@0.5: 27.3 -> 39.7 | Improved |
| use dual-granularity layout vs view-based baseline on Unique | dual-granularity framework · ScanRefer Unique | Acc@0.5: 68.9 (SeeGround) vs 66.8 | No clear effect |
| provide GT object classes | dual-granularity framework · Nr3D | overall accuracy: 63.2 -> 74.7 | Improved |
| use compact layout vs SPAZER agent | dual-granularity framework · Nr3D subset | inference time: 23.5 -> 10.9 s/query | Improved |
| use dual-granularity framework vs SPAZER | dual-granularity framework · ScanRefer subset | Acc@0.5: 48.8 -> 49.2 | No clear effect |
Closest papers by meaning
- Web-Scale Collection of Video Data for 4D Animal Reconstruction (NeurIPS 2025 · similarity 0.89)
- Reconstructing Animals and the Wild (CVPR 2025 · similarity 0.88)
- SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild (NeurIPS 2026 · similarity 0.88)
- WildBox: A Dataset and Benchmark for Aerial Monocular 3D Detection of African Savanna Wildlife (NeurIPS 2026 · similarity 0.87)
- BEAST3D: animal behavioral analysis and neural encoding from multi-view video via Gaussian splatting (NeurIPS 2026 · similarity 0.87)
- Learning the 3D Fauna of the Web (CVPR 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.
