Research map / 3D vision / Face and shape
Learning To Predict Scene-Level Implicit 3D From Posed RGBD Data
Nilesh Kulkarni, Linyi Jin, Justin Johnson, David F. Fouhey
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 |
|---|---|---|---|
| Use posed RGBD supervision instead of mesh supervision for DRDF training | D2-DRDF vs. Mesh DRDF · Matterport3D | Scene F1: 71.9 -> 72.1 | Improved |
| Use posed RGBD supervision instead of mesh supervision for DRDF training | D2-DRDF vs. Mesh DRDF · Matterport3D | Ray Occ. F1: 27.3 -> 25.1 | Hurt |
| Predict a DRDF rather than a density field | RGBD-supervised reconstruction · Matterport3D | Ray Occ. F1: 17.9 -> 25.1 | Improved |
| Use RGBD supervision rather than ODS mesh supervision with 50% of views | DRDF training · Matterport3D, 50% views | Scene F1: 68.4 -> 70.0 | Improved |
| Use RGBD supervision rather than ODS mesh supervision with 25% of views | DRDF training · Matterport3D, 25% views | Scene F1: 66.8 -> 70.0 | Improved |
| Use RGBD supervision rather than ODS mesh supervision with 3% of views | DRDF training · Omnidata, 3% views | Ray Occ. F1: 20.3 -> 26.7 | Improved |
| Add the sign entropy loss (ablation) | D2-DRDF · Scene reconstruction | Scene F1: 2 points | Improved |
| Adapt the pretrained full model using three auxiliary RGBD views and a reference view | D2-DRDF · Few-view adaptation | Ray Occ. F1: 11.2 -> 34.9 | Improved |
| Start adaptation from the depth-pretrained first-stage model rather than the full pretrained model (ablation) | D2-DRDF adaptation · Few-view adaptation | Ray Occ. F1: 34.9 -> 26.0 | Hurt |
| Adapt from scratch rather than starting with the full pretrained model (ablation) | D2-DRDF adaptation · Few-view adaptation | Ray Occ. F1: 34.9 -> 25.4 | Hurt |
Closest papers by meaning
- IPoD: Implicit Field Learning with Point Diffusion for Generalizable 3D Object Reconstruction from Single RGB-D Images (CVPR 2024 · similarity 0.87)
- DI-Fusion: Online Implicit 3D Reconstruction With Deep Priors (CVPR 2021 · similarity 0.86)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.86)
- O^2-Recon: Completing 3D Reconstruction of Occluded Objects in the Scene with a Pre-trained 2D Diffusion Model (AAAI 2024 · similarity 0.85)
- RGBD Objects in the Wild: Scaling Real-World 3D Object Learning from RGB-D Videos (CVPR 2024 · similarity 0.85)
- UniSDF: Unifying Neural Representations for High-Fidelity 3D Reconstruction of Complex Scenes with Reflections (NeurIPS 2024 · similarity 0.85)
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.
