Research map / 3D vision / Human pose estimation
Spatial Geometric Reasoning for Room Layout Estimation via Deep Reinforcement Learning
Liangliang Ren, Yangyang Song, Jiwen Lu, Jie Zhou
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 |
|---|---|---|---|
| make the feature network trainable instead of fixed (ablation) | LIM · LSUN validation | corner error (%): 6.53 -> 6.23 | Improved |
| make the feature network trainable instead of fixed (ablation) | LIM · LSUN validation | pixel error (%): 11.04 -> 10.00 | Improved |
| select the state of minimum critic value instead of the last state (10 steps) (ablation) | LIM · LSUN validation | corner error (%): 6.73 -> 6.42 | Improved |
| select the state of minimum critic value instead of the last state (10 steps) (ablation) | LIM · LSUN validation | pixel error (%): 10.65 -> 10.29 | Improved |
| select the state of minimum critic value instead of the last state (20 steps) (ablation) | LIM · LSUN validation | corner error (%): 7.65 -> 6.23 | Improved |
| select the state of minimum critic value instead of the last state (20 steps) (ablation) | LIM · LSUN validation | pixel error (%): 11.98 -> 10.00 | Improved |
| extend episode from 10 to 20 steps with last-state selection (ablation) | LIM · LSUN validation | corner error (%): 6.73 -> 7.65 | Hurt |
| extend episode from 10 to 20 steps with last-state selection (ablation) | LIM · LSUN validation | pixel error (%): 10.65 -> 11.98 | Hurt |
| extend episode from 10 to 20 steps with min-value selection (ablation) | LIM · LSUN validation | corner error (%): 6.42 -> 6.23 | Improved |
| extend episode from 10 to 20 steps with min-value selection (ablation) | LIM · LSUN validation | pixel error (%): 10.29 -> 10.00 | Improved |
| initialize from RoomNet predicted layout instead of typical layout (ablation) | LIM · LSUN validation | error | No clear effect |
| use multi-step RL layout adjustment instead of post-optimization (Delay) | Delay (2016) · LSUN validation speed | FPS: 0.03 -> 6.71 | 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.
- Delay (2016) — DeLay: Robust Spatial Layout Estimation for Cluttered Indoor Scenes (CVPR 2016) · better on 4 of 4 shared results
- LayoutNet (2018) — LayoutNet: Reconstructing the 3D Room Layout From a Single RGB Image (CVPR 2018) · better on 1 of 1 shared result
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
- GeoLayout: Geometry Driven Room Layout Estimation Based on Depth Maps of Planes (ECCV 2020 · similarity 0.90)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.88)
- OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization (NeurIPS 2025 · similarity 0.87)
- MetaSpatial: Reinforcing 3D Spatial Reasoning in VLMs for the Metaverse (ICLR 2026 · similarity 0.87)
- LayoutVLM: Differentiable Optimization of 3D Layout via Vision-Language Models (CVPR 2025 · similarity 0.87)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 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.
