Research map / Detection and segmentation / 3D detection and driving
RepVF: A Unified Vector Fields Representation for Multi-task 3D Perception
Jianbing Shen, Chunliang Li, Wencheng Han, Junbo Yin, Sanyuan Zhao
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 |
|---|---|---|---|
| use RepVF unified representation instead of task-specific representations (multi-head) (ablation) | multi-task 3D lane + object detection · OpenLane + Waymo 30% subset | F-Score: 58.7 -> 66.4 | Improved |
| use a single head instead of multiple heads with RepVF (ablation) | RFTR · OpenLane + Waymo 30% subset | Vehicle L1 mAP: 22.1 -> 25.3 | Improved |
| use a single head instead of multiple heads with RepVF (ablation) | RFTR · OpenLane + Waymo 30% subset | F-Score: 66.4 -> 66.5 | No clear effect |
| RepVF instead of anchor representation (ablation) | 3D lane detection · OpenLane (ablation) | F-Score: 53.4 -> 66.0 | Improved |
| RepVF instead of random point set (ablation) | 3D lane detection · OpenLane (ablation) | F-Score: 58.1 -> 66.0 | Improved |
| add simultaneous multi-task training with 3D object detection (ablation) | RepVF 3D lane detection · OpenLane (ablation) | F-Score: 66.0 -> 66.5 | Improved |
| use single-head RFTR instead of multi-head baseline (ablation) | multi-task training · gradient disparity | mean gradient disparity: 2.98 -> 2.47 | Improved |
| use multi-head with RepVF instead of multi-head with task-specific representations (ablation) | gradient disparity · gradient disparity | mean gradient disparity: 2.98 -> 2.72 | Improved |
| use lightweight multi-head with task-specific representation instead of larger head (ablation) | multi-head baseline · OpenLane + Waymo 30% subset | F-Score: 58.7 -> 49.6 | Hurt |
| use RFTR unified model instead of PETR | 3D object detection · Waymo val (L1) | pedestrian mAP: 12.1 -> 8.0 | Hurt |
| use RFTR unified model instead of PETR | 3D object detection · Waymo val (L1) | overall mAP: 20.9 -> 19.5 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- PETR — PETR: Position Embedding Transformation for Multi-View 3D Object Detection (ECCV 2022) · better on 1 of 4 shared results
- PersFormer — PersFormer: 3D Lane Detection via Perspective Transformer and the OpenLane Benchmark (ECCV 2022) · better on 1 of 1 shared result
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
- A Dynamic Feature Interaction Framework for Multi-task Visual Perception (IJCV 2023 · similarity 0.89)
- FULLER: Unified Multi-modality Multi-task 3D Perception via Multi-level Gradient Calibration (ICCV 2023 · similarity 0.88)
- M3Net: Multimodal Multi-task Learning for 3D Detection, Segmentation, and Occupancy Prediction in Autonomous Driving (AAAI 2025 · similarity 0.88)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.85)
- VGGDrive: Empowering Vision-Language Models with Cross-View Geometric Grounding for Autonomous Driving (CVPR 2026 · similarity 0.85)
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · 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.
