Research map / Detection and segmentation / 3D detection and driving
FrustumFormer: Adaptive Instance-Aware Resampling for Multi-View 3D Detection
Yuqi Wang, Yuntao Chen, Zhaoxiang Zhang
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 FrustumFormer instead of PolarFormer | nuScenes test with ResNet101-DCN · nuScenes test | mAP: 0.457 -> 0.478 | Improved |
| Use FrustumFormer instead of PETRv2 | nuScenes test with VoVNet-99 · nuScenes test | mAP: 0.490 -> 0.516 | Improved |
| Add instance frustum queries (ablation) | FrustumFormer baseline · nuScenes val ablation | mAP: 0.318 -> 0.326 | Improved |
| Add occupancy mask prediction (ablation) | FrustumFormer baseline · nuScenes val ablation | mAP: 0.318 -> 0.328 | Improved |
| Add instance frustum queries and occupancy mask (ablation) | FrustumFormer baseline · nuScenes val ablation | mAP: 0.318 -> 0.337 | Improved |
| Add temporal frustum fusion (ablation) | Adaptive instance-aware resampling · nuScenes val ablation | NDS: 0.383 -> 0.463 | Improved |
| Double sampling points uniformly across scene regions (ablation) | FrustumFormer baseline sampling · nuScenes val ablation | mAP: 0.318 -> 0.318 | No clear effect |
| Allocate extra sampling to instance regions rather than uniformly (ablation) | FrustumFormer baseline sampling · nuScenes val ablation | mAP: 0.318 -> 0.326 | Improved |
| Use strict projected boxes instead of enlarged boxes for occupancy supervision (ablation) | Occupancy mask learning · nuScenes val ablation | mAP: 0.324 -> 0.328 | Improved |
| Increase occupancy mask loss weight from 5.0 to 10.0 (ablation) | Occupancy mask learning · nuScenes val ablation | mAP: 0.328 -> 0.322 | Hurt |
| Decrease occupancy mask loss weight from 5.0 to 1.0 (ablation) | Occupancy mask learning · nuScenes val ablation | mAP: 0.328 -> 0.326 | Hurt |
| Increase temporal window from 8 to 16 with 4 keyframes (ablation) | Temporal frustum fusion · nuScenes val ablation | mAP: 0.360 -> 0.364 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PolarFormer — PolarFormer: Multi-Camera 3D Object Detection with Polar Transformer (AAAI 2023) · better on 4 of 4 shared results
- PETRv2 — PETRv2: A Unified Framework for 3D Perception from Multi-Camera Images (ICCV 2023) · better on 2 of 2 shared results
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
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.88)
- Instance-Aware Multi-Camera 3D Object Detection with Structural Priors Mining and Self-Boosting Learning (AAAI 2024 · similarity 0.86)
- BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-based Roadside 3D Object Detection (CVPR 2024 · similarity 0.86)
- GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025 · similarity 0.86)
- Far3D: Expanding the Horizon for Surround-View 3D Object Detection (AAAI 2024 · similarity 0.86)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 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.
