Research map / Point clouds / Segmentation and detection
TANet: Robust 3D Object Detection from Point Clouds with Triple Attention
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 TANet rather than PointPillars | KITTI 3D object detection · KITTI test split | 3D mAP (%): 60.80 -> 62.00 | Helped |
| Use TANet rather than PointRCNN | KITTI 3D object detection · KITTI test split | 3D mAP (%): 60.78 -> 62.00 | Helped |
| Use TANet rather than PointPillars under added noise | KITTI 3D detection, Pedestrians · KITTI validation, 100 noise points | 3D mAP: 55.04 -> 60.84 | Helped |
| Add point-wise attention alone (ablation) | TANet attention ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 65.59 -> 67.04 | Helped |
| Add channel-wise attention alone (ablation) | TANet attention ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 65.59 -> 66.93 | Helped |
| Combine point-wise and channel-wise attention in parallel (ablation) | TANet attention ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 65.59 -> 67.38 | Helped |
| Add voxel-wise attention to parallel point/channel attention (ablation) | TANet attention ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 67.38 -> 67.79 | Helped |
| Use PSA refinement without TA (ablation) | TANet regression ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 65.59 -> 66.21 | Helped |
| Use PSA with TA instead of TA alone (ablation) | TANet regression ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 67.79 -> 69.35 | Helped |
| Use PSA with TA rather than RefineDet with TA (ablation) | TANet regression ablation · KITTI validation, 100 noise points, all categories | 3D mAP: 68.80 -> 69.35 | Helped |
| Set the loss balance weight λ to 2.0 instead of 1.0 (ablation) | TANet · KITTI validation | 3D mAP: 70.69 -> 71.72 | Helped |
| Set the loss balance weight λ to 10.0 rather than 2.0 (ablation) | TANet · KITTI validation | 3D mAP: 71.72 -> 70.80 | Hurt |
Closest papers by meaning
- Uni3DETR: Unified 3D Detection Transformer (NeurIPS 2023 · similarity 0.89)
- 3D Object Detection Using Scale Invariant and Feature Reweighting Networks (AAAI 2019 · similarity 0.88)
- One for All: Multi-Domain Joint Training for Point Cloud Based 3D Object Detection (NeurIPS 2024 · similarity 0.86)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.85)
- PTT: Point-Trajectory Transformer for Efficient Temporal 3D Object Detection (CVPR 2024 · similarity 0.85)
- RobuRCDet: Enhancing Robustness of Radar-Camera Fusion in Bird's Eye View for 3D Object Detection (ICLR 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.