Research map / 3D vision / Gaussian splatting
GaussianOcc: Fully Self-supervised and Efficient 3D Occupancy Estimation with Gaussian Splatting
Wanshui Gan, Fang Liu, Hongbin Xu, Ningkai Mo, Naoto Yokoya
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 mask-out strategy to Gaussian splatting scale-aware loss (ablation) | GaussianOcc stage 1 · nuScenes depth | Abs Rel: 0.798 -> 0.293 | Improved |
| add erosion of the binary overlap mask (ablation) | GaussianOcc stage 1 · nuScenes depth | Abs Rel: 0.293 -> 0.281 | Improved |
| add depth refinement with fixed pose net (ablation) | GaussianOcc stage 1 · nuScenes depth | Abs Rel: 0.281 -> 0.258 | Improved |
| use two-stage training instead of one stage (ablation) | GaussianOcc · nuScenes depth | Abs Rel: 0.946 -> 0.211 | Improved |
| use learned pose instead of GT pose (ablation) | GaussianOcc stage 2 · nuScenes depth | Abs Rel: 0.214 -> 0.211 | No clear effect |
| use learned pose instead of prior-work learned pose (ablation) | GaussianOcc stage 2 · nuScenes depth | Abs Rel: 0.235 -> 0.211 | Improved |
| use learnable Gaussian scale instead of fixed 0.1 (ablation) | splatting rendering · nuScenes depth | Abs Rel: 0.217 -> 0.212 | Improved |
| use splatting rendering instead of volume rendering with GT pose (ablation) | GaussianOcc · nuScenes occupancy | mIoU*: 10.81 -> 11.30 | Improved |
| use splatting rendering instead of volume rendering with learned pose (ablation) | GaussianOcc · nuScenes occupancy | Abs Rel: 0.506 -> 0.197 | Improved |
| use splatting rendering with small scale 0.05 (ablation) | splatting rendering · nuScenes depth | Abs Rel: 0.215 -> 0.223 | Hurt |
| use splatting instead of volume rendering for render time | GaussianOcc · nuScenes occupancy | render time per 180x320 (s): 0.85 -> 0.17 | Improved |
| use splatting instead of volume rendering for training time | GaussianOcc · nuScenes occupancy | training time per epoch (h): 2.68 -> 1 | 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.
- R3D3 — R3D3: Dense 3D Reconstruction of Dynamic Scenes from Multiple Cameras (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
- GaussTR: Foundation Model-Aligned Gaussian Transformer for Self-Supervised 3D Spatial Understanding (CVPR 2025 · similarity 0.89)
- ODG: Occupancy Prediction Using Dual Gaussians (NeurIPS 2025 · similarity 0.89)
- VoxelSplat: Dynamic Gaussian Splatting as an Effective Loss for Occupancy and Flow Prediction (CVPR 2025 · similarity 0.88)
- SelfSplat: Pose-Free and 3D Prior-Free Generalizable 3D Gaussian Splatting (CVPR 2025 · similarity 0.88)
- FreeSplat: Generalizable 3D Gaussian Splatting Towards Free View Synthesis of Indoor Scenes (NeurIPS 2024 · similarity 0.88)
- GaussianFormer-2: Probabilistic Gaussian Superposition for Efficient 3D Occupancy Prediction (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.
