Research map / Detection and segmentation / 3D detection and driving
ScanRefer: 3D Object Localization in RGB-D Scans using Natural Language
Dave Zhenyu Chen, Angel X. Chang, Matthias Nießner
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 |
|---|---|---|---|
| train detection backbone jointly with localization (end-to-end) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 25.72 -> 27.40 | Improved |
| use 3D end-to-end model instead of 2D One-stage + backprojection | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 9.04 -> 27.40 | Improved |
| add language-to-object classifier (xyz+multiview+normals) (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 26.90 -> 27.40 | Improved |
| add language-to-object classifier (xyz) (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 23.76 -> 24.22 | Mixed |
| add language-to-object classifier (xyz+rgb) (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 23.61 -> 24.32 | Improved |
| add language-to-object classifier (xyz+multiview) (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 26.23 -> 26.50 | Hurt |
| replace raw RGB with multiview image features (with lobjcls) (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 24.32 -> 26.50 | Improved |
| add normals to xyz+rgb (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 23.61 -> 25.21 | Improved |
| add normals to xyz+multiview (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 26.23 -> 26.90 | Improved |
| add RGB colors to xyz (with lobjcls) (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 24.22 -> 24.32 | Improved |
| feed whole descriptions instead of first sentences (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 26.12 -> 27.40 | Improved |
| feed whole descriptions instead of object names (ablation) | ScanRefer · ScanRefer val overall | Acc@0.5IoU: 25.05 -> 27.40 | Improved |
Closest papers by meaning
- Multi3DRefer: Grounding Text Description to Multiple 3D Objects (ICCV 2023 · similarity 0.92)
- Scan2Cap: Context-Aware Dense Captioning in RGB-D Scans (CVPR 2021 · similarity 0.89)
- WildRefer: 3D Object Localization in Large-scale Dynamic Scenes with Multi-modal Visual Data and Natural Language (ECCV 2024 · similarity 0.89)
- Vote2Cap-DETR++: Decoupling Localization and Describing for End-to-End 3D Dense Captioning (TPAMI 2024 · similarity 0.88)
- LOCATE 3D: Real-World Object Localization via Self-Supervised Learning in 3D (ICML 2025 · similarity 0.86)
- AugRefer: Advancing 3D Visual Grounding via Cross-Modal Augmentation and Spatial Relation-based Referring (AAAI 2025 · 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.
