Research map / Detection and segmentation / 3D detection and driving
CAT-Det: Contrastively Augmented Transformer for Multi-Modal 3D Object Detection
Yanan Zhang, Jiaxin Chen, Di Huang
Check 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 CAT-Det's two-stream transformer, cross-modal fusion, and OMDA | Multi-modal 3D detection · KITTI test split | mAP: 65.19 -> 67.05 | Improved |
| Use CAT-Det rather than Part-A² | Multi-modal 3D detection · KITTI validation split | mAP: 72.20 -> 75.42 | Improved |
| Replace PointNet++ and 2D CNN with TPI while retaining concatenation (ablation) | Baseline multi-modal detector · KITTI validation split | mAP: 68.65 -> 70.44 | Improved |
| Add CMT to TPI (ablation) | TPI multi-modal detector · KITTI validation split | mAP: 70.44 -> 71.89 | Improved |
| Add OMDA to TPI + CMT (ablation) | TPI + CMT detector · KITTI validation split | mAP: 71.89 -> 75.42 | Improved |
| Use CMT fusion instead of concatenation or attention-based fusion (ablation) | TPI multi-modal detector · KITTI validation split | mAP: 70.44 -> 71.89 | Improved |
| Preserve single-modal features alongside cross-modal features in CMT (ablation) | CMT fusion · KITTI validation split | mAP: 71.41 -> 71.89 | Improved |
| Add point-level contrastive augmentation (ablation) | TPI + CMT detector · KITTI validation split | mAP: 71.89 -> 74.54 | Improved |
| Add object-level contrastive augmentation with background negatives (ablation) | Point-level contrastive augmentation · KITTI validation split | mAP: 74.54 -> 75.00 | Improved |
| Select object-level negative pairs using a memory bank (ablation) | Point-level contrastive augmentation · KITTI validation split | mAP: 74.54 -> 75.42 | Improved |
| Retain CMT at each of Layers 1–5 (ablation) | CAT-Det · KITTI validation split | mAP: 75.15 -> 75.42; 74.86 -> 75.42; 74.42 -> 75.42; 74.55 -> 75.42; 74.13 -> 75.42 | Improved |
| Set the OMDA memory bank size to 1024 rather than 256, 512, 2048, or 4096 (ablation) | CAT-Det OMDA · KITTI validation split | mAP: 75.14; 75.19; 75.42; 75.28; 75.06 | Mixed |
Closest papers by meaning
- Cross Modal Transformer: Towards Fast and Robust 3D Object Detection (ICCV 2023 · similarity 0.89)
- MV2DFusion: Leveraging Modality-Specific Object Semantics for Multi-Modal 3D Detection (TPAMI 2025 · similarity 0.88)
- Unleash the Potential of Image Branch for Cross-modal 3D Object Detection (NeurIPS 2023 · similarity 0.88)
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.88)
- CALICO: Self-Supervised Camera-LiDAR Contrastive Pre-training for BEV Perception (ICLR 2024 · similarity 0.86)
- AS-Det: Active Sampling for Adaptive 3D Object Detection in Point Clouds (AAAI 2025 · 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.
