FreeLoc: Online Floorplan Localization via Diffusion-Aided Pose Refinement

Haocheng Peng; Boyang Zhou; Jiarui Hu; Xiyue Guo; Ziyang Zhang; Boming Zhao; Yifan Gao; Xiao Li

cs.ROOct 4, 2026

FreeLoc: Online Floorplan Localization via Diffusion-Aided Pose Refinement

Authors: Haocheng Peng, Boyang Zhou, Jiarui Hu, Xiyue Guo, Ziyang Zhang, Boming Zhao, Yifan Gao, Xiao Li, +2 more

Organizations: State Key Lab of CAD&CG, Zhejiang University

Abstract

Floorplans provide compact and widely available geometric maps for indoor localization, but existing high-performing floorplan-based methods still convert them into dense scene-specific offline databases, tying accuracy, storage, and runtime to the sampling resolution of the discretized pose space. We present FreeLoc, an online RGB-based floorplan localization framework that treats the floorplan as a directly queryable geometric map. FreeLoc introduces an efficient online geometric querying and diffusion-aided refinement scheme, which retrieves plausible pose anchors through on-the-fly floorplan ray querying and refines them into accurate continuous pose estimates. For sequential localization, FreeLoc develops an online likelihood construction strategy that bridges single-frame localization and probabilistic temporal fusion by constructing likelihoods from coarse-sampled candidates and refined pose hypotheses, enabling histogram-filter-based temporal fusion without offline databases. Experiments demonstrate real-time online inference and state-of-the-art performance in both single-frame and sequential localization, while real-world results validate practical deployability in indoor robotic localization scenarios.

Figures & tables

Figure 1: FreeLoc enables database-free online floorplan localization, directly using the provided floorplan to support accurate single-frame and sequential localization in new environments.

Figure 2: Overview of FreeLoc. We retrieve top- $K$ coarse anchors by matching image-side rays with online queried floorplan rays, then refine and re-score them for the final estimate. For sequential localization, coarse and refined candidates construct online likelihoods for temporal fusion.

Method	Gibson(f)				Structured3D
Method	$@0.1\mathrm{m}\uparrow$	$@0.5\mathrm{m}\uparrow$	$@1\mathrm{m}\uparrow$	$@1\mathrm{m}30^{\circ}\uparrow$	$@0.1\mathrm{m}\uparrow$	$@0.5\mathrm{m}\uparrow$	$@1\mathrm{m}\uparrow$	$@1\mathrm{m}30^{\circ}\uparrow$
PF-Net	0	1.0	4.1	1.1	0.1	1.2	4.4	1.4
LASER	0.2	3.9	9.8	6.2	0.5	5.2	10.0	7.8
F 3 Loc	5.3	30.7	37.9	36.3	1.7	15.0	23.0	21.9
Ours	10.0	39.4	44.6	42.9	3.1	23.9	30.5	29.1

Table 1: Single-frame localization recall (%) on Gibson(f) and Structured3D. Bold denotes the best result.

Method	$\mathrm{SR}@1\mathrm{m}$ (%) $\uparrow$	$\mathrm{RMSE}_{\mathrm{succ}}$ (m) $\downarrow$	$\mathrm{RMSE}_{\mathrm{all}}$ (m) $\downarrow$
PF-Net	8.0	0.43	4.27
LASER	35.0	0.31	2.69
F 3 Loc	86.5	0.14	0.87
Ours	100.0	0.15	0.15

Table 2: Sequential localization results on Gibson(t).

Figure 3: Real-world performance of FreeLoc. The center panel shows the floorplan with converged predicted and ground-truth trajectories; surrounding panels show time-ordered RGB observations and posterior probabilities. Blue indicates predictions and green indicates ground truth.

Initial error	Before (m)	After (m)
$\leq 0.5\,\mathrm{m}$	0.25	0.18
$\leq 1.0\,\mathrm{m}$	0.35	0.25
$\leq 1.5\,\mathrm{m}$	0.42	0.32
All	3.14	3.09

Table 3: Pose refinement under different coarse-initialization quality.

Refinement	SR@1m (%) $\uparrow$	SR@0.5m (%) $\uparrow$	SR@0.2m (%) $\uparrow$	RMSE all (m) $\downarrow$
ICP Optimization	97.3	97.3	43.2	0.21
Diffusion-Aided Pose Refinement	100.0	100.0	51.4	0.15

Table 4: Sequential localization comparison between ICP optimization and diffusion-aided pose refinement on Gibson(t).

Setting	Level	SR@1m	RMSE succ	RMSE all
Clean	–	100.0	0.145	0.145
Gaussian	Mild	100.0	0.147	0.147
	Moderate	100.0	0.171	0.171
	Severe	97.3	0.235	0.307
Motion-dep.	Mild	100.0	0.145	0.145
	Moderate	100.0	0.150	0.150

Table 5: Robustness to inter-frame ego-motion uncertainty on Gibson(t).

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 4: Qualitative results of diffusion-aided pose refinement. Starting from coarse pose anchors retrieved by online floorplan sampling, our diffusion-aided pose refinement module progressively predicts residual pose corrections guided by paired image-side and floorplan-side ray geometry. The color transition from light blue to dark blue indicates the pose recovery process, while green denotes the ground truth. Coarse anchors from different initial locations are refined toward poses closer to the ground truth, validating the effectiveness of diffusion-aided pose refinement.

Figure 5: Qualitative comparison of single-frame localization results. We compare FreeLoc with representative baselines on Gibson and Structured3D. Compared with PF-Net, LASER, and F 3 Loc, FreeLoc produces pose estimates that are better aligned with the ground truth, benefiting from accurate online coarse pose retrieval and diffusion-aided pose refinement.

Figure 6: Qualitative comparison of sequential localization results. We compare FreeLoc with representative baselines on the Gibson(t) sequential localization dataset. The results show the temporal evolution of posterior distributions. Benefiting from more reliable single-frame localization evidence, FreeLoc concentrates the posterior around the correct region more quickly and maintains more accurate localization over time.

Refinement	SR@1m (%) $\uparrow$	SR@0.5m (%) $\uparrow$	SR@0.2m (%) $\uparrow$	RMSE all (m) $\downarrow$
Direct Regression	97.3	97.3	48.6	0.20
Diffusion-Aided Pose Refinement	100.0	100.0	51.4	0.15

Appendix

Table 7: Ablation study of diffusion-aided pose refinement against direct regression on Gibson(t).

Method	Gibson(f)				Structured3D
Method	$@0.1\mathrm{m}\uparrow$	$@0.5\mathrm{m}\uparrow$	$@1\mathrm{m}\uparrow$	$@1\mathrm{m}30^{\circ}\uparrow$	$@0.1\mathrm{m}\uparrow$	$@0.5\mathrm{m}\uparrow$	$@1\mathrm{m}\uparrow$	$@1\mathrm{m}30^{\circ}\uparrow$
UnLoc	13.5	46.1	49.6	47.8	3.6	27.7	33.6	32.6
Ours	16.6	46.9	50.2	48.7	4.9	29.5	34.3	33.0

Appendix

Table 8: Controlled comparison with UnLoc under matched depth predictions. FreeLoc uses the same UnLoc depth predictions at test time without retraining the diffusion refinement model. Bold denotes the better result.

Explore similar work

From Uncertainty to Determinism: Coarse-to-Fine Visual Floorplan Localization without Ray Matching
Jul 29, 2026Shiyong Meng, Bolei Chen, Ping Zhong +4Indoor Localization Uncertainty
School of Computer Science and Engineering, Central South University
SceneAligner: 3D-Grounded Floorplan Localization in the Wild
May 21, 2026Junhyeong Cho, Ruojin Cai, Hadar Averbuch-ElorFloor Plan Generation Indoor Localization
Cornell University · Kempner Institute, Harvard University
GALoc: Gravity Aligned Wireframes for Depth-Free Monocular Floorplan Localization
Sep 8, 2026Jeahn Han, Minji Kim, Jeongbin Sohn +3Indoor Localization Floor Plan Generation
GIST · KAIST · Zurich University of Applied Sciences

Jul 29, 2026, cs.ROJ/K move · Enter open · S save

From Uncertainty to Determinism: Coarse-to-Fine Visual Floorplan Localization without Ray Matching

Shiyong Meng, Bolei Chen, Ping Zhong, Yang Wan+3

School of Computer Science and Engineering, Central South University

Indoor Localization Uncertainty

Open paper AskPDF arXiv

Visual Floorplan Localization (FLoc) has emerged as a promising solution for indoor localization by matching egocentric images against minimalist structural maps. However, due to cross-modal information asymmetry and repetitive indoor layouts, visual FLoc is fundamentally challenged by multimodal pose distributions, where visually identical observations map to distinct, spatially separated locations. Existing ray-matching-based methods tackle this by explicitly predicting sparse geometric or semantic rays, which inherently incur information loss and demand resource-intensive preprocessing alongside exhaustive matching during inference. In this paper, we bypass the intermediate ray-matching paradigm and propose a coarse-to-fine visual FLoc framework that progresses from uncertainty to determinism. In the coarse stage, we design an image-conditioned pose diffusion model to parameterize the continuous multimodal pose distribution, effectively routing stochastically initialized pose particles toward distinct candidate modes. In the refinement stage, we propose a localized refiner that predicts bounded sub-meter pose residuals from candidate-centered floorplan crops, where structural ambiguities are largely eliminated. Our method effectively balances global multi-hypothesis tracking and local sub-meter refinement without requiring any offline map preprocessing or test-time lookup tables. Comprehensive results on the S3D (full) and ZInD benchmarks demonstrate that our approach achieves state-of-the-art accuracy and robustness.

Abstract

Explore similar work

From Uncertainty to Determinism: Coarse-to-Fine Visual Floorplan Localization without Ray Matching

SceneAligner: 3D-Grounded Floorplan Localization in the Wild

GALoc: Gravity Aligned Wireframes for Depth-Free Monocular Floorplan Localization