Research map / Detection and segmentation / 3D detection and driving
Scene Reconstruction as Mapping Priors for 3D Detection
Yang Fu, Yuliang Zou, Hao Xiang, Xin Huang, Yijing Bai, Chen Song, Weijing Shi, Govind Thattai, Dragomir Anguelov, Mingxing Tan, Yingwei Li
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 surfel and 3DGS mapping priors (ablation) | Ours-baseline · WOD validation subset | Overall L2 APH: 80.1 -> 81.7 | Improved |
| add surfel and 3DGS mapping priors (ablation) | SWFormer-dagger · WOD validation subset | Overall L2 APH: 71.6 -> 72.7 | Improved |
| add surfel prior only (ablation) | Ours-baseline · WOD validation subset | Overall L2 APH: 80.1 -> 81.1 | Improved |
| add 3DGS prior only (ablation) | Ours-baseline · WOD validation subset | Overall L2 APH: 80.1 -> 81.0 | Improved |
| add surfel or 3DGS prior to weaker baseline (ablation) | SWFormer-dagger · WOD validation subset | Pedestrian L2 AP: 73.8 -> 73.5 | Hurt |
| use gated fusion instead of concatenation (ablation) | MPA3D fusion · WOD validation subset | Overall L2 AP: 80.4 -> 83.3 | Improved |
| use gated fusion instead of averaging (ablation) | MPA3D fusion · WOD validation subset | Overall L2 APH: 77.0 -> 81.7 | Improved |
| use gated fusion instead of sum (ablation) | MPA3D fusion · WOD validation subset | Overall L2 APH: 73.4 -> 81.7 | Improved |
| add camera input to LiDAR (ablation) | MPA3D-96M · WOD validation | Overall L2 APH: 74.9 -> 75.7 | Improved |
| add surfel prior to LiDAR+camera (ablation) | MPA3D-96M · WOD validation | Overall L2 APH: 75.7 -> 76.1 | Improved |
| add 3DGS prior to LiDAR+camera+surfel (ablation) | MPA3D-96M · WOD validation | Overall L2 APH: 76.1 -> 77.4 | Improved |
| use mapping priors with 4 frames instead of 100-frame temporal fusion | MPA3D vs MAD · WOD test leaderboard | Overall L2 APH: 80.2 -> 81.6 | Improved |
Closest papers by meaning
- PriorDrive: Enhancing Online HD Mapping with Unified Vector Priors (AAAI 2026 · similarity 0.87)
- R3D3: Dense 3D Reconstruction of Dynamic Scenes from Multiple Cameras (ICCV 2023 · similarity 0.87)
- Map-Det3D: Metric Feed-Forward 3D Reconstruction Prior for Multi-view 3D Object Detection from Streaming Inputs (ECCV 2026 · similarity 0.86)
- Pseudo-LiDAR From Visual Depth Estimation: Bridging the Gap in 3D Object Detection for Autonomous Driving (CVPR 2019 · similarity 0.86)
- MonoPlace3D: Learning 3D-Aware Object Placement for 3D Monocular Detection (CVPR 2025 · similarity 0.86)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 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.
