Research map / 3D vision / 3D generation
Occupancy as Set of Points
Yiang Shi, Tianheng Cheng, Qian Zhang, Wenyu Liu, Xinggang Wang
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 |
|---|---|---|---|
| add decoder layers (1 to 3) (ablation) | OSP · Occ3D-nuScenes | mIoU: 36.76 -> 38.01 | Improved |
| add adaptive oversampling during training (ablation) | OSP · Occ3D-nuScenes | mIoU: 38.01 -> 38.48 | Improved |
| add random point perturbation (0.1m) to grid centers (ablation) | OSP · Occ3D-nuScenes | mIoU: 38.48 -> 38.67 | Improved |
| increase multi-scale image feature maps from 2 to 4 (ablation) | OSP · Occ3D-nuScenes | mIoU: 38.67 -> 39.41 | Improved |
| add dice loss (ablation) | OSP · Occ3D-nuScenes | mIoU: 38.88 -> 39.41 | Improved |
| use OSP as plugin refiner for BEVFormer over full 80m x 80m scene | BEVFormer · Occ3D-nuScenes | mIoU: 37.84 -> 41.21 | Improved |
| increase refine scale from 20m to 80m (ablation) | BEVFormer w/ OSP · Occ3D-nuScenes | mIoU: 38.20 -> 41.21 | Improved |
| adaptive sampling of uncertain points at inference (ablation) | OSP · Occ3D-nuScenes | mIoU: 39.41 -> 39.08 | Mixed |
| adaptive sampling of uncertain points for BEVFormer refinement (ablation) | BEVFormer w/ OSP · Occ3D-nuScenes | mIoU: 39.35 -> 39.22 | Mixed |
| evaluate points outside the training range (60m-80m) | OSP · Occ3D-nuScenes | mIoU: 42.29 -> 23.00 | Hurt |
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.
- SurroundOcc — SurroundOcc: Multi-camera 3D Occupancy Prediction for Autonomous Driving (ICCV 2023) · better on 1 of 1 shared result
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
- AdaOcc: Adaptive 3D Occupancy Prediction for Embodied Tasks (NeurIPS 2026 · similarity 0.89)
- Spatiotemporal Decoupling for Efficient Vision-Based Occupancy Forecasting (CVPR 2025 · similarity 0.87)
- RadOcc: Learning Cross-Modality Occupancy Knowledge through Rendering Assisted Distillation (AAAI 2024 · similarity 0.87)
- COTR: Compact Occupancy TRansformer for Vision-based 3D Occupancy Prediction (CVPR 2024 · similarity 0.87)
- Learning Occupancy for Monocular 3D Object Detection (CVPR 2024 · similarity 0.87)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 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.
