Research map / Detection and segmentation / 3D detection and driving
ViewFormer: Exploring Spatiotemporal Modeling for Multi-View 3D Occupancy Perception via View-Guided Transformers
Jinke Li, Xiao He, Chonghua Zhou, Xiaoqiang Cheng, Yang Wen, Dan 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 |
|---|---|---|---|
| replace projection-first deformable attention with learning-first view attention (ablation) | ViewFormer · FlowOcc3D (no flow head) | mIoU: 42.39 -> 43.61 | Improved |
| learn sample points in a query-specific view coordinate system instead of ego-centric coordinates (ablation) | view attention · FlowOcc3D (no flow head) | mIoU: 42.86 -> 43.61 | Improved |
| add streaming temporal attention (ablation) | ViewFormer · FlowOcc3D (with flow head) | mIoU: 39.28 -> 42.54 | Improved |
| compress voxel queries to BEV queries for temporal interaction (BEV-to-BEV vs Voxel-to-BEV) (ablation) | streaming temporal attention · FlowOcc3D (with flow head) | mIoU: 41.62 -> 42.54 | Improved |
| voxel-to-BEV temporal interaction flow convergence (ablation) | streaming temporal attention · FlowOcc3D (with flow head) | mAVE: 0.412 -> 1.104 | Hurt |
| increase streaming memory queue length from 1 to 3 frames (ablation) | streaming temporal attention · FlowOcc3D (with flow head) | mIoU: 41.67 -> 42.42 | Improved |
| streaming memory attention adds FPS cost (ablation) | ViewFormer · FlowOcc3D | FPS: 4.2 -> 4.1 | No clear effect |
| train temporal attention with streaming video instead of sliding window (ablation) | ViewFormer · FlowOcc3D | mIoU: 42.17 -> 42.54; 36 h -> 13 h | Improved |
| use deformable attention instead of CNN-based temporal interaction (ablation) | streaming temporal attention · FlowOcc3D | mIoU: 41.95 -> 42.54 | Improved |
| apply view attention to MapTR | MapTR · nuScenes HD map | mAP: 44.73 -> 50.55 | Improved |
| apply view attention to DETR3D | DETR3D · nuScenes detection | mAP: 0.347 -> 0.388 | Improved |
| supervise flow with occupancy-level instead of object-level flow | ViewFormer · FlowOcc3D (qualitative) | flow quality | 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.
- BEVFormer — BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (ECCV 2022) · better on 4 of 4 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 Flexible 3D Perception: Object-Centric Occupancy Completion Augments 3D Object Detection (NeurIPS 2024 · similarity 0.89)
- Spatiotemporal Decoupling for Efficient Vision-Based Occupancy Forecasting (CVPR 2025 · similarity 0.88)
- COTR: Compact Occupancy TRansformer for Vision-based 3D Occupancy Prediction (CVPR 2024 · similarity 0.88)
- STCOcc: Sparse Spatial-Temporal Cascade Renovation for 3D Occupancy and Scene Flow Prediction (CVPR 2025 · similarity 0.87)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.87)
- 3D Occupancy Prediction with Low-Resolution Queries via Prototype-aware View Transformation (CVPR 2025 · similarity 0.87)
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.
