Research map / 3D vision / 3D generation
Learning 3D Scene Priors With 2D Supervision
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace the permutation-invariant transformer with an LSTM (ablation) | 3D scene prior · 3D-FRONT bedroom and living room | FID: 21.59 -> 67.97; 40.47 -> 86.77 | Hurt |
| use a hypersphere latent space rather than uniform latent parameterization (ablation) | 3D scene prior · 3D-FRONT bedroom and living room | FID: 30.61 -> 21.59; 95.51 -> 40.47 | Helped |
| train layout first, then jointly train layout and shape instead of one-stage training (ablation) | 3D scene prior · 3D-FRONT bedroom and living room | FID: 24.24 -> 21.59; 43.43 -> 40.47 | Helped |
| increase training view coverage from 25% to full (ablation) | 3D scene prior · 3D-FRONT bedroom and living room | FID: 29.48 -> 21.59; 105.33 -> 40.47 | Helped |
| increase training view coverage from 50% to full (ablation) | 3D scene prior · 3D-FRONT bedroom and living room | FID: 24.91 -> 21.59; 44.85 -> 40.47 | Helped |
| add shape retrieval post-processing (ablation) | single-view reconstruction · ScanNet | CD: 1.01 -> 0.79 | Helped |
| add shape retrieval post-processing (ablation) | single-view reconstruction · ScanNet | 3D Box IoU: 0.25 -> 0.27 | Helped |
Closest papers by meaning
- Learning 3D Object Shape and Layout Without 3D Supervision (CVPR 2022 · similarity 0.93)
- 3DP3: 3D Scene Perception via Probabilistic Programming (NeurIPS 2021 · similarity 0.89)
- 3D-RCNN: Instance-Level 3D Object Reconstruction via Render-and-Compare (CVPR 2018 · similarity 0.89)
- BlockGAN: Learning 3D Object-aware Scene Representations from Unlabelled Images (NeurIPS 2020 · similarity 0.89)
- Neural Part Priors: Learning To Optimize Part-Based Object Completion in RGB-D Scans (CVPR 2023 · similarity 0.89)
- Cooperative Holistic Scene Understanding: Unifying 3D Object, Layout, and Camera Pose Estimation (NeurIPS 2018 · 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.