Research map / 3D vision / 3D generation
SDFormer: Vision-based 3D Semantic Scene Completion via SAM-assisted Dual-channel Voxel Transformer
Yujie Xue, Huilong Pi, Jiapeng Zhang, Yunchuan Qin, Zhuo Tang, Kenli Li, Ruihui 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 |
|---|---|---|---|
| add Semantic Constructor (SAM-based semantic voxel) to geometric voxel baseline (ablation) | SDFormer · SemanticKITTI val | mIoU: 15.43 -> 16.12 | Improved |
| use SAM guidance in Geometric Refiner (ablation) | SDFormer · SemanticKITTI val | IoU: 43.85 -> 45.10 | Improved |
| add Learnable Mask Generation (SAM shape prior) (ablation) | SDFormer · SemanticKITTI val | mIoU: +0.81 mIoU, +0.39 IoU | Improved |
| add Shape Segments Interactive (SSI) (ablation) | SDFormer · SemanticKITTI val | mIoU: 16.58 -> 17.18 | Improved |
| add Semantic Calibration Affinity (SCA) (ablation) | SDFormer · SemanticKITTI val | mIoU: 16.43 -> 17.18 | Improved |
| use SSI instead of attention fusion (ablation) | SDFormer · SemanticKITTI val | mIoU: 16.43 -> 17.18 | Improved |
| use SSI instead of add-conv fusion (ablation) | SDFormer · SemanticKITTI val | mIoU: 15.94 -> 17.18 | Improved |
| use SDFormer vs CGFormer on 3D SSC at 51.2m | SemanticKITTI val · SemanticKITTI val (51.2m) | IoU: 45.99 -> 45.65 | Hurt |
| use SDFormer vs multi-frame HTCL-S at 12.8m | SemanticKITTI val · SemanticKITTI val (12.8m) | mIoU: 27.37 -> 26.45 | Hurt |
| use SDFormer vs CGFormer on KITTI-360 | SSCBench-KITTI-360 test · SSCBench-KITTI-360 test | mIoU: 20.05 -> 20.59 | Improved |
Closest papers by meaning
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · similarity 0.93)
- H2GFormer: Horizontal-to-Global Voxel Transformer for 3D Semantic Scene Completion (AAAI 2024 · similarity 0.92)
- Memory-Augmented Re-Completion for 3D Semantic Scene Completion (AAAI 2025 · similarity 0.91)
- Voxel Proposal Network via Multi-Frame Knowledge Distillation for Semantic Scene Completion (NeurIPS 2024 · similarity 0.90)
- IPFormer: Visual 3D Panoptic Scene Completion with Context-Adaptive Instance Proposals (NeurIPS 2025 · similarity 0.90)
- Not All Voxels Are Equal: Hardness-Aware Semantic Scene Completion with Self-Distillation (CVPR 2024 · similarity 0.89)
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.
