Research map / 3D vision / 3D generation
Cascaded Context Pyramid for Full-Resolution 3D Semantic Scene Completion
Pingping Zhang, Wei Liu, Yinjie Lei, Huchuan Lu, Xiaoyun 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 |
|---|---|---|---|
| add Cascaded Context Pyramid (CCP) instead of PPM (ablation) | CCPNet · SUNCG (ablation) | SC IoU: decrease 4.1% with PPM; 87.3 vs 91.4 | Improved |
| add CCP instead of ASPP (ablation) | CCPNet · SUNCG (ablation) | SC IoU: decrease 2.3% with ASPP; 89.1 vs 91.4 | Improved |
| use Guided Residual Block instead of basic residual blocks (ablation) | CCPNet · SUNCG (ablation) | SC and SSC IoU: decreasing 8.1% and 8.4% for SC and SSC | Improved |
| add guidance to Guided Residual Block (ablation) | CCPNet · SUNCG (ablation) | SC/SSC IoU: 84.3 -> 91.4 SC; 67.4 -> 74.2 SSC | Improved |
| add feature amplification to Guided Residual Block (ablation) | CCPNet · SUNCG (ablation) | SC/SSC IoU: 88.7 -> 91.4 SC; 73.6 -> 74.2 SSC | Improved |
| predict at full resolution instead of quarter or half resolution (ablation) | CCPNet · SUNCG (ablation) | SC and SSC IoU: improving the IoUs by 4.9% and 5.1% | Improved |
| replace 3D dilated convolutions with separated kernels (ablation) | SSCNet · SUNCG | SC/SSC IoU and cost: 73.5 -> 76.8 SC; 46.4 -> 52.5 SSC; params 930 -> 532 | Improved |
| use CCPNet instead of SATNet | 3D semantic scene completion · NYU Kinect | SSC average IoU: 34.4 -> 41.3 | Improved |
| add SUNCG pretraining data to NYU training | CCPNet · NYU Kinect | SSC average IoU: 38.5 -> 41.3 | Improved |
| use CCPNet instead of DDRNet | 3D semantic scene completion · NYU Kinect efficiency | speed and parameters: 658 ms -> 57 ms; 195k -> 89k params | 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.
- DDRNet — DDRNet: Depth Map Denoising and Refinement for Consumer Depth Cameras Using Cascaded CNNs (ECCV 2018) · better on 4 of 4 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
- Semantic Scene Completion From a Single Depth Image (CVPR 2017 · similarity 0.94)
- CasFusionNet: A Cascaded Network for Point Cloud Semantic Scene Completion by Dense Feature Fusion (AAAI 2023 · similarity 0.91)
- Context and Geometry Aware Voxel Transformer for Semantic Scene Completion (NeurIPS 2024 · similarity 0.90)
- Memory-Augmented Re-Completion for 3D Semantic Scene Completion (AAAI 2025 · similarity 0.89)
- H2GFormer: Horizontal-to-Global Voxel Transformer for 3D Semantic Scene Completion (AAAI 2024 · similarity 0.88)
- Bi-SSC: Geometric-Semantic Bidirectional Fusion for Camera-based 3D Semantic Scene Completion (CVPR 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.
