Research map / Detection and segmentation / 3D detection and driving
RECO: Region-Aware Compensation for Extrinsic Perturbations in Roadside 3D Detection
Junsheng Du, Yuhuan Lu, Zhaocheng He
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 region-aware compensation (soft) instead of none, yaw mu=0 (ablation) | RECO · DAIR-V2X-I persistent yaw | Car AP easy: 67.5595 -> 70.4639 | Improved |
| add region-aware compensation (soft) instead of none, z mu=0 (ablation) | RECO · DAIR-V2X-I persistent z | Car AP easy: 48.3313 -> 72.616 | Improved |
| replace global compensation with soft range-aware compensation, yaw mu=0 (ablation) | RECO · DAIR-V2X-I persistent yaw | Car AP easy: 33.2416 -> 70.4639 | Improved |
| use soft sigmoid gate instead of hard assignment, yaw mu=-2 (ablation) | RECO · DAIR-V2X-I persistent yaw | Car AP easy: 40.0773 -> 49.3382 | Improved |
| use soft gate instead of hard assignment, z mu=0 (ablation) | RECO · DAIR-V2X-I persistent z | Car AP easy: 68.3201 -> 72.616 | Improved |
| use soft gate instead of hard assignment, z mu=-2 (ablation) | RECO · DAIR-V2X-I persistent z | Car AP easy: 31.6210 -> 31.0973 | Mixed |
| add global compensation to uncompensated model, yaw mu=0 (ablation) | detector · DAIR-V2X-I persistent yaw | Car AP easy: 67.5595 -> 33.2416 | Hurt |
| add global compensation to uncompensated model, yaw mu=-2 (ablation) | detector · DAIR-V2X-I persistent yaw | Car AP easy: 18.0282 -> 32.2572 | Improved |
| compare RECO with BEVHeight++ under z transient perturbation | RECO · DAIR-V2X-I (z) | Car AP easy: 55.8025 -> 72.6160 | Improved |
| compare RECO with BEVHeight++ under Rope3D yaw | RECO · Rope3D (yaw) | Car AP easy: 50.3703 -> 65.3942 | Improved |
| compare RECO with BEVHeight under pitch perturbation | RECO · DAIR-V2X-I (pitch) | Car AP easy: 53.1881 -> 65.3235 | Improved |
| compare RECO with BEVHeight on pedestrian under pitch | RECO · DAIR-V2X-I (pitch) | Pedestrian AP easy: 9.3655 -> 8.4951 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- BEVHeight — BEVHeight: A Robust Framework for Vision-Based Roadside 3D Object Detection (CVPR 2023) · better on 13 of 13 shared results
- BEVSpread — BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-based Roadside 3D Object Detection (CVPR 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learnability-Driven Submodular Optimization for Active Roadside 3D Detection (CVPR 2026 · similarity 0.88)
- BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-based Roadside 3D Object Detection (CVPR 2024 · similarity 0.87)
- Monocular 3D Object Detection: An Extrinsic Parameter Free Approach (CVPR 2021 · similarity 0.87)
- GSV2X: Geometry-Aware Uncertainty Modeling and Orthogonal Fusion for Robust Roadside Perception (CVPR 2026 · similarity 0.87)
- RESOLVE: A Multi-Resolution and Multi-Modal Dataset for Roadside Cooperative Perception (ECCV 2026 · similarity 0.87)
- BEVHeight: A Robust Framework for Vision-Based Roadside 3D Object Detection (CVPR 2023 · 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.
