Research map / 3D vision / Face and shape
3D Scene Reconstruction from a Single Viewport
Maximilian Denninger, Rudolph Triebel
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 |
|---|---|---|---|
| remove loss shaping (ablation) | Tree net · SUNCG (GT normals) | IOU: 65.10 -> 70.49 | Hurt |
| add loss shaping (ablation) | Tree net · SUNCG (GT normals) | Hausdorff distance (lower better): 0.0745 -> 0.0416 | Improved |
| add loss shaping (ablation) | Tree net · Replica (GT normals) | Hausdorff distance (lower better): 0.0614 -> 0.0387 | Improved |
| add loss shaping (ablation) | Tree net · Replica (generated normals) | Hausdorff distance (lower better): 0.0766 -> 0.0562 | Improved |
| add loss shaping (ablation) | Tree net · SUNCG (generated normals) | IOU: 67.61 -> 61.56 | Mixed |
| reduce tree height from 5 to 4 (ablation) | Tree net · SUNCG (GT normals) | IOU: 65.10 -> 68.05 | Improved |
| reduce tree height from 5 to 4 (ablation) | Tree net · Replica (GT normals) | IOU: 73.97 -> 78.76 | Improved |
| increase tree height from 5 to 6 (ablation) | Tree net · SUNCG (GT normals) | Hausdorff distance (lower better): 0.0416 -> 0.0390 | Improved |
| increase tree height from 5 to 6 (ablation) | Tree net · Replica (GT normals) | Hausdorff distance (lower better): 0.0387 -> 0.0457 | Hurt |
| use generated U-Net normals instead of GT normals | Tree net default · SUNCG | IOU: 65.10 -> 61.56 | Hurt |
| use generated U-Net normals instead of GT normals | Tree net default · Replica | IOU: 73.97 -> 65.86 | Hurt |
| evaluate on real Replica instead of synthetic SUNCG (no finetuning) | Tree net default · Replica vs SUNCG | IOU: 65.10 -> 73.97 | Improved |
Closest papers by meaning
- CoReNet: Coherent 3D Scene Reconstruction from a Single RGB Image (ECCV 2020 · similarity 0.90)
- Atlas: End-to-End 3D Scene Reconstruction from Posed Images (ECCV 2020 · similarity 0.89)
- VolumeFusion: Deep Depth Fusion for 3D Scene Reconstruction (ICCV 2021 · similarity 0.89)
- Monocular Scene Reconstruction with 3D SDF Transformers (ICLR 2023 · similarity 0.89)
- NeuralRecon: Real-Time Coherent 3D Reconstruction From Monocular Video (CVPR 2021 · similarity 0.88)
- NeuralRecon: Real-Time Coherent 3D Scene Reconstruction From Monocular Video (TPAMI 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.
