Research map / 3D vision / Human pose estimation
LaLaLoc: Latent Layout Localisation in Dynamic, Unvisited Environments
Henry Howard-Jenkins, Jose-Raul Ruiz-Sarmiento, Victor Adrian Prisacariu
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 latent pose optimisation after Vogel disc re-sampling (ablation) | LaLaLoc · Structured3D refinement | median pose error (cm): 11.0 -> 8.3 | Improved |
| add Vogel disc re-sampling before latent pose optimisation (ablation) | LaLaLoc · Structured3D refinement | median pose error (cm): 10.5 -> 8.3 | Improved |
| optimise in latent space instead of decoding layouts explicitly (ablation) | LaLaLoc · Structured3D refinement | median pose error (cm): 12.2 -> 8.3 | Improved |
| use Chamfer distance on point clouds as layout similarity (ablation) | LaLaLoc · Structured3D val | layout recall@1: 80.3% -> 90.8% | Improved |
| train with 2 stages instead of end-to-end (ablation) | LaLaLoc · Structured3D val (RGB query) | layout recall@1: 24.4% -> 70.7% | Improved |
| add decoder loss to log-ratio loss for layout branch (ablation) | LaLaLoc · Structured3D val | pose recall@1: 77.3% -> 87.6% | Improved |
| add log-ratio loss to decoder loss for layout branch (ablation) | LaLaLoc · Structured3D val | layout recall@1: 90.9% -> 89.3% | No clear effect |
| add cross-modal log-ratio loss to L2 for image branch (ablation) | LaLaLoc · Structured3D val | layout recall@1: 70.7% -> 60.7% | Hurt |
| use ResNet50 instead of ResNet18 for layout branch (ablation) | LaLaLoc · Structured3D val | layout recall@1: 89.3% -> 87.5% | Hurt |
| use 2-layer MLP instead of FC predictor (ablation) | LaLaLoc · Structured3D val | layout recall@1: 89.3% -> 87.2% | Hurt |
| increase furniture level from empty to full | 2D ICP · Structured3D | median pose error (cm): 2.4 -> 23.0 | Hurt |
| increase furniture level from empty to full | LaLaLoc · Structured3D | median pose error (cm): 5.7 -> 8.3 | Mixed |
Later papers that compared with this method
2 later accepted papers report LaLaLoc as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- LaLaLoc++: Global Floor Plan Comprehension for Layout Localisation in Unvisited Environments (ECCV 2022 · better on 11 of 11 shared results)
- LASER: LAtent SpacE Rendering for 2D Visual Localization (CVPR 2022 · 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
- LaLaLoc++: Global Floor Plan Comprehension for Layout Localisation in Unvisited Environments (ECCV 2022 · similarity 0.93)
- SPVLoc: Semantic Panoramic Viewport Matching for 6D Camera Localization in Unseen Environments (ECCV 2024 · similarity 0.87)
- LayoutNet: Reconstructing the 3D Room Layout From a Single RGB Image (CVPR 2018 · similarity 0.86)
- MvFFN: Multi-view Floor-Plan Feed-Forward Network for Unposed Wide-Baseline Panorama Layout Reconstruction (NeurIPS 2026 · similarity 0.85)
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.84)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.84)
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.
