Research map / Detection and segmentation / 3D detection and driving
Cyclic Refiner: Object-Aware Temporal Representation Learning for Multi-view 3D Detection and Tracking
Mingzhe Guo, Zhipeng Zhang, Liping Jing, Yuan He, Ke Wang, Heng Fan
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 cyclic refiner on image features (ablation) | BEVFormer-Small · nuScenes detection val | mAP: 0.370 -> 0.386 | Improved |
| add cyclic refiner on BEV features (ablation) | BEVFormer-Small · nuScenes detection val | mAP: 0.370 -> 0.390 | Improved |
| refine both image and BEV features (ablation) | BEVFormer-Small · nuScenes detection val | mAP: 0.370 -> 0.398 | Improved |
| add temporal fusion on image features without backward refinement (ablation) | Cycer-S · nuScenes detection val | mAP: 0.370 -> 0.361 | Hurt |
| add backward refinement together with image temporal fusion (ablation) | Cycer-S · nuScenes detection val | mAP: 0.361 -> 0.398 | Improved |
| add object-aware association (OAA) to tracker (ablation) | Cycer-S · nuScenes tracking val | AMOTA: 0.305 -> 0.397 | Improved |
| add OAA to baseline BEVFormer-S detector (ablation) | BEVFormer-S · nuScenes tracking val | AMOTA: 0.274 -> 0.337 | Improved |
| add multi-clue matching (ablation) | Cycer-S · nuScenes tracking val | AMOTA: 0.305 -> 0.356 | Improved |
| add buffering strategy (ablation) | Cycer-S · nuScenes tracking val | AMOTA: 0.305 -> 0.349 | Improved |
| add cascaded scale-aware matching (ablation) | Cycer-S · nuScenes tracking val | AMOTA: 0.305 -> 0.365 | Improved |
| use Cyclic Refiner instead of Ocean object-aware strategy (ablation) | Cycer-S · nuScenes detection val | mAP: 0.367 -> 0.398 | Improved |
| set image scale levels to 3 instead of 5 (ablation) | Cycer-S · nuScenes detection val | mAP: 0.383 -> 0.386 | Improved |
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.
- BEVFormer-Small — BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (ECCV 2022) · better on 6 of 6 shared results
- SparseBEV — SparseBEV: High-Performance Sparse 3D Object Detection from Multi-Camera Videos (ICCV 2023) · 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
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024 · similarity 0.88)
- Looking Beyond Two Frames: End-to-End Multi-Object Tracking Using Spatial and Temporal Transformers (TPAMI 2022 · similarity 0.88)
- Time Will Tell: New Outlooks and A Baseline for Temporal Multi-View 3D Object Detection (ICLR 2023 · similarity 0.87)
- Temporal Coherent Object Flow for Multi-Object Tracking (AAAI 2025 · similarity 0.86)
- SynCL: A Synergistic Training Strategy with Instance-Aware Contrastive Learning for End-to-End Multi-Camera 3D Tracking (NeurIPS 2025 · similarity 0.86)
- S2-Track: A Simple yet Strong Approach for End-to-End 3D Multi-Object Tracking (ICML 2025 · similarity 0.85)
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.
