Research map / 3D vision / 3D generation
ScanQA: 3D Question Answering for Spatial Scene Understanding
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use ScanQA's joint 3D-language model with object localization and object classification | ScanQA · ScanQA test w/ objects | EM@10: 50.76 -> 56.51 | Helped |
| Add object classification to answer prediction (ablation) | ScanQA · ScanQA test w/ objects | EM@1: 12.16 -> 18.31 | Helped |
| Add object localization to answer prediction (ablation) | ScanQA · ScanQA test w/ objects | EM@1: 12.16 -> 20.46 | Helped |
| Add object classification to answer prediction when localization is present (ablation) | ScanQA · ScanQA test w/ objects | EM@1: 20.46 -> 23.45 | Helped |
| Add object localization when object classification is present (ablation) | ScanQA · ScanQA test w/ objects | EM@1: 18.31 -> 23.45 | Helped |
| Add RGB values to xyz features (ablation) | ScanQA · ScanQA test w/ objects | Acc@0.25: 23.67 -> 23.45 | Hurt |
| Add multiview image features to xyz features (ablation) | ScanQA · ScanQA test w/ objects | Acc@0.25: 23.67 -> 25.40 | Helped |
| Add normals to xyz+multiview features (ablation) | ScanQA · ScanQA test w/ objects | EM@10: 55.51 -> 56.51 | Helped |
| Predict multiple objects rather than a single object (ablation) | ScanQA · ScanQA test w/ objects | EM@10: 56.51 -> 55.99 | Hurt |
| Set transformer depth to L=1 rather than L=2 (ablation) | ScanQA · ScanQA test w/ objects | EM@1: 23.45 -> 23.83 | Helped |
| Increase transformer depth to L=3 (ablation) | ScanQA · ScanQA test w/ objects | EM@10: 56.51 -> 44.71 | Hurt |
| Set hidden size to d=128 rather than d=256 (ablation) | ScanQA · ScanQA test w/ objects | EM@10: 56.51 -> 56.71 | Helped |
Closest papers by meaning
- 3D Question Answering via only 2D Vision-Language Models (ICML 2025 · similarity 0.92)
- Bridging the Gap between 2D and 3D Visual Question Answering: A Fusion Approach for 3D VQA (AAAI 2024 · similarity 0.92)
- Chat-Scene: Bridging 3D Scene and Large Language Models with Object Identifiers (NeurIPS 2024 · similarity 0.90)
- Hypo3D: Exploring Hypothetical Reasoning in 3D (ICML 2025 · similarity 0.89)
- SpatiaLQA: A Benchmark for Evaluating Spatial Logical Reasoning in Vision-Language Models (CVPR 2026 · similarity 0.89)
- Zero-shot 3D Question Answering via Voxel-based Dynamic Token Compression (CVPR 2025 · similarity 0.88)
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.