Research map / 3D vision / 3D generation
Robust 3D Semantic Occupancy Prediction With Calibration-Free Spatial Transformation
Zhuangwei Zhuang, Ziyin Wang, Sitao Chen, Lizhao Liu, Hui Luo, Mingkui Tan
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 |
|---|---|---|---|
| add image feature aggregation (IFA) (ablation) | REO baseline · Occ3D-nuScenes val | mIoU (%): 42.40 -> 43.04 | Improved |
| add BEV contextual block (BCB) (ablation) | REO with IFA · Occ3D-nuScenes val | IoU (%): 80.81 -> 81.97 | Improved |
| add BEV contextual block (BCB) without auxiliary tasks (ablation) | REO with IFA · Occ3D-nuScenes val | mIoU (%): 43.04 -> 42.68 | Hurt |
| add 2D/3D auxiliary training tasks (ablation) | REO with IFA and BCB · Occ3D-nuScenes val | mIoU (%): 42.68 -> 46.22 | Improved |
| fuse only 2560 point cloud voxels instead of camera-only (ablation) | REO · SemanticKITTI val | SC IoU (%): 35.34 -> 51.95 | Improved |
| fuse 20480 instead of 2560 point cloud voxels (ablation) | REO · SemanticKITTI val | SC IoU (%): 51.95 -> 56.56 | Improved |
| add calibration noise to input (ablation) | REO · SemanticKITTI val | SC IoU (%): 56.56 -> 56.06 | No clear effect |
| use only 4th-stage image features instead of all stages (ablation) | REO · Occ3D-nuScenes val | mIoU (%): 46.21 -> 46.22 | No clear effect |
| add focal loss weight 1.0 (ablation) | REO · Occ3D-nuScenes val | mIoU (%): 44.25 -> 46.22 | Improved |
| add dice loss weight 1.0 (ablation) | REO · Occ3D-nuScenes val | mIoU (%): 42.72 -> 46.22 | Improved |
| add depth regression task (weight 2.0 vs 0) (ablation) | REO · Occ3D-nuScenes val | mIoU (%): 45.94 -> 46.51 | Improved |
| add texture reconstruction task (weight 1.0 vs 0) (ablation) | REO · Occ3D-nuScenes val | mIoU (%): 45.85 -> 46.51 | 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.
- COTR — COTR: Correspondence Transformer for Matching Across Images (ICCV 2021) · 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
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · similarity 0.92)
- CSV-Occ: Fusing Multi-frame Alignment for Occupancy Prediction with Temporal Cross State Space Model and Central Voting Mechanism (ICML 2025 · similarity 0.91)
- OctOcc: High-Resolution 3D Occupancy Prediction with Octree (AAAI 2024 · similarity 0.91)
- SDGOCC: Semantic and Depth-Guided Bird's-Eye View Transformation for 3D Multimodal Occupancy Prediction (CVPR 2025 · similarity 0.91)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.90)
- AdaOcc: Adaptive 3D Occupancy Prediction for Embodied Tasks (NeurIPS 2026 · similarity 0.90)
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.
