Research map / Detection and segmentation / 3D detection and driving
VGGT-Det: Mining VGGT Internal Priors for Sensor-Geometry-Free Multi-View Indoor 3D Object Detection
Yang Cao, Feize Wu, Dave Zhenyu Chen, Yingji Zhong, Lanqing Hong, Dan Xu
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 Attention-Guided Query Generation (replace FPS query sampling) (ablation) | VGGT-Det basic backbone · ScanNet (SG-Free) | mAP@0.25: 41.4 -> 44.2 | Improved |
| add Query-Driven Feature Aggregation (See-Query) (ablation) | VGGT-Det backbone+AG · ScanNet (SG-Free) | mAP@0.25: 44.2 -> 46.9 | Improved |
| use multi-level VGGT layers (4/11/17/23) instead of last layer only (ablation) | VGGT-Det decoder · ScanNet (SG-Free) | mAP@0.25: 37.5 -> 44.2 | Improved |
| query-driven adaptive aggregation instead of sequential layer feeding (ablation) | VGGT-Det decoder · ScanNet (SG-Free) | mAP@0.25: 44.2 -> 46.9 | Improved |
| increase eval input frames from 40 to 80 (ablation) | VGGT-Det · ScanNet | mAP@0.25: 45.3 -> 46.9 | Improved |
| VGGT-Det instead of MVSDet with VGGT poses | SG-Free detector · ARKitScenes (SG-Free) | mAP@0.25: 19.4 -> 28.0 | Improved |
| VGGT-Det instead of MVSDet with VGGT poses | SG-Free detector · ScanNet (SG-Free) | mAP@0.25: 42.5 -> 46.9 | Improved |
| VGGT-Det instead of MVSDet in memory | VGGT pipeline · ScanNet (40 frames) | extra memory (GB): 13.81 -> 3.57 | Improved |
| VGGT-Det instead of MVSDet in time | VGGT pipeline · ScanNet (40 frames) | extra time (s): 0.21 -> 0.23 | No clear effect |
Closest papers by meaning
- VGGT: Visual Geometry Grounded Transformer (CVPR 2025 · similarity 0.90)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.87)
- V-DETR: DETR with Vertex Relative Position Encoding for 3D Object Detection (ICLR 2024 · similarity 0.87)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.87)
- UniDet3D: Multi-dataset Indoor 3D Object Detection (AAAI 2025 · similarity 0.87)
- MVSDet: Multi-View Indoor 3D Object Detection via Efficient Plane Sweeps (NeurIPS 2024 · 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.
