Research map / Point clouds / Segmentation and detection
Unleash the Potential of Image Branch for Cross-modal 3D Object Detection
Yifan Zhang, Qijian Zhang, Junhui Hou, Yixuan Yuan, Guoliang Xing
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 bidirectional propagation to single-modal point branch (ablation) | UPIDet · KITTI val | mAP: 75.88 -> 78.64 | Improved |
| add point-to-pixel propagation used only in training (ablation) | single-modal UPIDet · KITTI val | mAP: 75.88 -> 77.01 | Improved |
| add pixel-to-point propagation only (ablation) | single-modal UPIDet · KITTI val | mAP: 75.88 -> 77.47 | Improved |
| add point-to-pixel on top of pixel-to-point (ablation) | UPIDet · KITTI val | mAP: 77.47 -> 78.64 | Improved |
| add 3D segmentation and center estimation auxiliary tasks (ablation) | LiDAR-only branch · KITTI val | mAP: 75.18 -> 75.88 | Improved |
| add 3D segmentation alone (ablation) | LiDAR-only branch · KITTI val | mAP: 75.18 -> 75.07 | No clear effect |
| add image branch without 2D auxiliary tasks (ablation) | UPIDet · KITTI val | mAP: 75.88 -> 76.38 | Improved |
| add 2D semantic segmentation (ablation) | UPIDet with image branch · KITTI val | mAP: 76.38 -> 77.18 | Improved |
| add NLC map estimation (ablation) | UPIDet with 2D seg · KITTI val | mAP: 77.18 -> 78.64 | Improved |
| apply bidirectional propagation to SASA asymmetric backbone | SASA · KITTI val | mAP: 75.24 -> 76.61 | Improved |
| use image semantics in semantic-guided SA sampling (ablation) | UPIDet · KITTI SA Level-2 | foreground rate: 15.87 -> 20.70 | Improved |
| joint 2D-3D learning vs single-modal 2D segmentation | PSPNet · KITTI 2D segmentation | mIoU: 43.92 -> 48.34 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PointAugmenting — PointAugmenting: Cross-Modal Augmentation for 3D Object Detection (CVPR 2021) · 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
- Bidirectional Propagation for Cross-Modal 3D Object Detection (ICLR 2023 · similarity 0.90)
- Cross Modal Transformer: Towards Fast and Robust 3D Object Detection (ICCV 2023 · similarity 0.90)
- PointAugmenting: Cross-Modal Augmentation for 3D Object Detection (CVPR 2021 · similarity 0.88)
- CAT-Det: Contrastively Augmented Transformer for Multi-Modal 3D Object Detection (CVPR 2022 · similarity 0.88)
- CCF: Complementary Collaborative Fusion for Domain Generalized Multi-Modal 3D Object Detection (CVPR 2026 · similarity 0.88)
- SimDistill: Simulated Multi-Modal Distillation for BEV 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.
