Research map / 3D vision / 3D generation
Progressive Gaussian Transformer with Anisotropy-aware Sampling for Open Vocabulary Occupancy Prediction
Chi Yan, Dan Xu
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 progressive online densification (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 14.84 -> 15.15 | Improved |
| add anisotropy-aware feature sampling (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 15.03 -> 15.15 | Improved |
| add asymmetric self-attention (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 14.85 -> 15.15 | Improved |
| add progressive online densification (ablation) | PG-Occ · nuScenes retrieval | mAP (v): 19.21 -> 21.20 | Improved |
| add 1000 extended queries in the progressive layer (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 14.84 -> 15.15 | Improved |
| increase extended queries from 1000 to 2000 (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 15.15 -> 14.79 | Mixed |
| increase base queries from 1000 to 4000 (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 13.17 -> 15.15 | Improved |
| increase base queries from 4000 to 8000 (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 15.15 -> 14.99 | Hurt |
| lower densification threshold from 0.2 to 0.0 (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 15.15 -> 15.11 | No clear effect |
| increase sampling offsets from 8 to 32 (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 15.05 -> 15.46 | Improved |
| add cosine feature loss (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 13.51 -> 15.15 | Improved |
| add temporal photometric loss (ablation) | PG-Occ · Occ3D-nuScenes | mIoU: 14.47 -> 15.15 | Improved |
Closest papers by meaning
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.91)
- GaussianFormer-2: Probabilistic Gaussian Superposition for Efficient 3D Occupancy Prediction (CVPR 2025 · similarity 0.91)
- GaussTR: Foundation Model-Aligned Gaussian Transformer for Self-Supervised 3D Spatial Understanding (CVPR 2025 · similarity 0.90)
- GaussianWorld: Gaussian World Model for Streaming 3D Occupancy Prediction (CVPR 2025 · similarity 0.89)
- OctOcc: High-Resolution 3D Occupancy Prediction with Octree (AAAI 2024 · similarity 0.88)
- VoxelSplat: Dynamic Gaussian Splatting as an Effective Loss for Occupancy and Flow Prediction (CVPR 2025 · similarity 0.88)
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.
