Research map / 3D vision / 3D generation
SPSC: Sparse and Scalable Multi-Modal 3D Occupancy Prediction for Autonomous Driving
Qingju Guo, Shuang Li, Binhui Xie, Jing Geng, Wei Li
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 |
|---|---|---|---|
| use SPSC instead of OccMamba (multi-modal) | SPSC · nuScenes-Occupancy val (multi-modal) | mIoU (%): 25.1 -> 28.1 | Improved |
| use SPSC instead of OccGen (camera-only) | C-SPSC · nuScenes-Occupancy val (camera-only) | mIoU (%): 14.5 -> 15.2 | Improved |
| use SPSC instead of OccMamba (LiDAR-only) | L-SPSC · nuScenes-Occupancy val (LiDAR-only) | mIoU (%): 22.7 -> 25.2 | Improved |
| add multiple historical frames as input | SPSC · nuScenes-Occupancy val (multi-modal) | mIoU (%): 28.1 -> 30.1 | Improved |
| use LiDAR-guided image sampling instead of BEVFormer sampling (ablation) | query initialization · nuScenes-Occupancy val (multi-modal) | mIoU (%): 26.8 -> 28.1 | Improved |
| use LiDAR-guided image sampling instead of LSS sampling (ablation) | query initialization · nuScenes-Occupancy val (multi-modal) | mIoU (%): 27.6 -> 28.1 | Improved |
| use ground-truth-guided sampling instead of LiDAR-guided (ablation) | query initialization · nuScenes-Occupancy val (multi-modal) | mIoU (%): 28.1 -> 28.4 | Improved |
| increase feature interaction scale from 64x64x5 to 512x512x40 (ablation) | SPSC · nuScenes-Occupancy val (multi-modal) | mIoU (%): 20.4 -> 28.1 | Improved |
| combine Z-order and Hilbert serialization randomly per layer (ablation) | query serialization · nuScenes-Occupancy val (multi-modal) | mIoU (%): 27.5 -> 28.1 | Improved |
| increase top-K final-layer queries from 480k to 560k (ablation) | SPSC · nuScenes-Occupancy val (multi-modal) | mIoU (%): 28.1 -> 27.8 | Hurt |
| enlarge sliding window from 1024 to 2048 (ablation) | SPSC · nuScenes-Occupancy val (multi-modal) | mIoU (%): 28.1 -> 28.0 | Hurt |
| use attention blocks instead of sparse convolution (ablation) | SPSC · nuScenes-Occupancy val (multi-modal) | mIoU (%): 24.3 -> 28.1 | 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.
- OccGen — OccGen: Generative Multi-modal 3D Occupancy Prediction for Autonomous Driving (ECCV 2024) · 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
- 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 0.91)
- SparseOcc: Rethinking Sparse Latent Representation for Vision-Based Semantic Occupancy Prediction (CVPR 2024 · similarity 0.91)
- OPUS: Occupancy Prediction Using a Sparse Set (NeurIPS 2024 · 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.
