Research map / 3D vision / 3D generation
Global-Aware Monocular Semantic Scene Completion with State Space Models
Shijie Li, Zhongyao Cheng, Rong Li, Shuai Li, Juergen Gall, Xun Xu, Xulei Yang
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 |
|---|---|---|---|
| replace 3D conv blocks with Mamba blocks in the decoder (ablation) | GA-MonoSSC · Occ-ScanNet-mini | mIoU: 46.83 -> 48.19 | Improved |
| replace 3D conv blocks with Mamba blocks in the decoder (ablation) | GA-MonoSSC · NYUv2 | mIoU: 31.6 -> 32.32 | Improved |
| add frustum reordering of voxels before Mamba (ablation) | GA-MonoSSC · Occ-ScanNet-mini | mIoU: 46.66 -> 48.19 | Improved |
| add frustum reordering of voxels before Mamba (ablation) | GA-MonoSSC · NYUv2 | IoU: 47.1 -> 47.51 | Improved |
| add multi-scale feature learning (ablation) | GA-MonoSSC · Occ-ScanNet-mini | mIoU: 47.39 -> 48.19 | Improved |
| learn semantic and geometric features separately (multi-modality) (ablation) | GA-MonoSSC · Occ-ScanNet-mini | mIoU: 46.2 -> 48.19 | Improved |
| learn multi-modality and multi-scale features together (ablation) | GA-MonoSSC · Occ-ScanNet-mini | mIoU: 46.07 -> 48.19 | Improved |
| supervise geometry decoder with 3D occupancy instead of 2D depth (ablation) | GA-MonoSSC · NYUv2 | mIoU: 31.65 -> 32.32 | Improved |
| inject geometry features into semantic decoder (geo->sem) vs none (ablation) | GA-MonoSSC · NYUv2 | mIoU: 32.0 -> 32.32 | Improved |
| inject semantic features into geometry decoder (sem->geo) (ablation) | GA-MonoSSC · NYUv2 | mIoU: 32.0 -> 31.61 | Hurt |
| make geometry and semantic decoders interact in both directions (ablation) | GA-MonoSSC · NYUv2 | mIoU: 32.32 -> 32.07 | No clear effect |
| use GA-MonoSSC instead of ISO | GA-MonoSSC · Occ-ScanNet | mIoU: 28.71 -> 35.65 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MonoScene — MonoScene: Monocular 3D Semantic Scene Completion (CVPR 2022) · better on 4 of 4 shared results
- NDC-Scene — NDC-Scene: Boost Monocular 3D Semantic Scene Completion in Normalized Device Coordinates Space (ICCV 2023) · better on 2 of 2 shared results
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
- MonoScene: Monocular 3D Semantic Scene Completion (CVPR 2022 · similarity 0.91)
- H2GFormer: Horizontal-to-Global Voxel Transformer for 3D Semantic Scene Completion (AAAI 2024 · similarity 0.89)
- IPFormer: Visual 3D Panoptic Scene Completion with Context-Adaptive Instance Proposals (NeurIPS 2025 · similarity 0.89)
- Memory-Augmented Re-Completion for 3D Semantic Scene Completion (AAAI 2025 · similarity 0.89)
- Skip Mamba Diffusion for Monocular 3D Semantic Scene Completion (AAAI 2025 · similarity 0.88)
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · 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.
