Research map / 3D vision / Face and shape
CIRCLE: Convolutional Implicit Reconstruction and Completion for Large-Scale Indoor Scene
Hao-Xiang Chen, Jiahui Huang, Tai-Jiang Mu, Shi-Min Hu
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 inference-time differentiable-rendering refinement (ablation) | CIRCLE · N-Matterport3D | F-score (%): 89.11 -> 89.23 | Improved |
| add inference-time differentiable-rendering refinement (ablation) | CIRCLE · N-Matterport3D | RMSE (x1e-3): 16.5 -> 16.2 | Improved |
| add inference-time differentiable-rendering refinement (ablation) | CIRCLE · ICL-NUIM (zero-shot) | F-score (%): 23.89 -> 25.54 | Improved |
| add inference-time differentiable-rendering refinement (ablation) | CIRCLE · ICL-NUIM (zero-shot) | RMSE (x1e-3): 22.7 -> 22.1 | Improved |
| use implicit differentiation instead of unrolled auto-differentiation (Ours-AD) in the renderer (ablation) | CIRCLE · N-Matterport3D with noisy poses | reconstruction quality | Improved |
| use implicit differentiation instead of unrolled auto-differentiation (Ours-AD) in the renderer (ablation) | CIRCLE · rendering 300,000 rays | GPU memory: 4.6G -> 1.3G | Improved |
| add a small normal loss weight (alpha in [0,1]) (ablation) | CIRCLE · N-Matterport3D | precision | Mixed |
| decrease voxel size from 10 cm to 5 cm (ablation) | CIRCLE · N-Matterport3D | reconstruction quality and recall | Improved |
| train with larger voxel size, test at 7.5 cm (ablation) | CIRCLE · N-Matterport3D | CD: test CD approx 0.019 for both 5 cm and 10 cm | No clear effect |
| use sparse convolutional U-Net representation over dense baselines | CIRCLE · run time | inference time: 10-50x faster than ConvON and SPSG | 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.
- DI-Fusion — DI-Fusion: Online Implicit 3D Reconstruction With Deep Priors (CVPR 2021) · better on 6 of 7 shared results
- SPSG — SPSG: Self-Supervised Photometric Scene Generation From RGB-D Scans (CVPR 2021) · better on 6 of 6 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
- Learning Long-Range Information with Dual-Scale Transformers for Indoor Scene Completion (ICCV 2023 · similarity 0.88)
- Semantic Scene Completion via Integrating Instances and Scene In-the-Loop (CVPR 2021 · similarity 0.88)
- Convolutional Occupancy Networks (ECCV 2020 · similarity 0.88)
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.86)
- Behind the Veil: Enhanced Indoor 3D Scene Reconstruction with Occluded Surfaces Completion (CVPR 2024 · similarity 0.86)
- Point-based Instance Completion with Scene Constraints (ICLR 2025 · similarity 0.86)
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.
