Research map / Detection and segmentation / 3D detection and driving
VISTA: Boosting 3D Object Detection via Dual Cross-VIew SpaTial Attention
Check 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 |
|---|---|---|---|
| Apply VISTA multi-view fusion to OHS | OHS · nuScenes validation | mAP: 59.5 -> 60.8 (+1.3) | Helped |
| Apply VISTA multi-view fusion to CenterPoint | CenterPoint · nuScenes validation | NDS: 64.8 -> 65.6 (+0.8) | Helped |
| Use average fusion instead of adaptive cross-view attention (ablation) | OHS multi-view fusion · nuScenes validation | mAP: 59.5 -> 59.2 | Hurt |
| Replace convolutional attention with linear attention (ablation) | VISTA · nuScenes validation | mAP: 60.0 -> 58.7 | Hurt |
| Add convolutional attention instead of direct baseline fusion (ablation) | OHS · nuScenes validation | mAP: 59.5 -> 60.0 | Helped |
| Add attention variance constraint (ablation) | VISTA with convolutional attention · nuScenes validation | mAP: 60.0 -> 60.4 | Helped |
| Add decoupled classification and regression attention (ablation) | VISTA with variance constraint · nuScenes validation | mAP: 60.4 -> 60.8 | Helped |
| Apply the variance constraint to classification and regression branches rather than only one branch (ablation) | Decoupled VISTA · nuScenes validation | mAP: 60.6 and 60.5 -> 60.8 | Helped |
| Replace decoupling with a deeper single-attention head (ablation) | VISTA · nuScenes validation | mAP: 60.40 -> 60.45 (+0.05) | No clear effect |
| Use attention variance loss instead of background prediction scaling (ablation) | VISTA · nuScenes validation | mAP: 59.6; 60.4 -> 60.8 | Helped |
| Apply VISTA fusion to CenterPoint | CenterPoint · Waymo test LEVEL_2 Pedestrian | mAPH: 71.5 -> 72.5 | Helped |
| Apply VISTA fusion to CenterPoint | CenterPoint · Waymo test LEVEL_2 Cyclist | mAPH: 71.2 -> 71.6 | Helped |
Closest papers by meaning
- Multi-View 3D Object Detection Network for Autonomous Driving (CVPR 2017 · similarity 0.91)
- Fully Convolutional One-Stage 3D Object Detection on LiDAR Range Images (NeurIPS 2022 · similarity 0.90)
- Fully Sparse 3D Object Detection (NeurIPS 2022 · similarity 0.90)
- A Versatile Multi-View Framework for LiDAR-Based 3D Object Detection With Guidance From Panoptic Segmentation (CVPR 2022 · similarity 0.90)
- Towards Unsupervised Object Detection From LiDAR Point Clouds (CVPR 2023 · similarity 0.90)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.89)
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.