Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Perspective from a Higher Dimension: Can 3D Geometric Priors Help Visual Floorplan Localization?

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Injecting 3D geometric priors learned from RGB-D scans into a floorplan localizer's observation model lifts single-frame success at 0.1 m from 1.5% to 5.6% on Structured3D and from 8.3% to 9.4% on Gibson(g), with no added inference cost.

desk verdict Solid gains in floorplan localization, but the causal claim that 3D geometric priors drive them is underdetermined by the experiments as run. read the letter →

arxiv 2507.18881 v1 pith:YC66K7KC submitted 2025-07-25 cs.CV cs.RO

classification cs.CVcs.RO
keywords FloorplanLocalizationGeometry-ConstrainedViewInvarianceView-SceneAlignedGeometricPriorContrastiveLearning3DPriorsSelf-SupervisedPretrainingVisualRGB-DSceneUnderstanding
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Visual floorplan localization asks a camera to find its pose on a 2D floorplan from RGB images alone, despite the gap between sparse geometric maps and appearance-rich photos. This paper argues that the missing link is 3D geometry: if the visual encoder is first taught, through self-supervised contrastive learning on ScanNet RGB-D scans, what stays invariant across views of the same 3D points and how views align with 3D surface geometry, the resulting representation yields better localization once fine-tuned in the F3Loc framework. On Structured3D, single-frame success at 0.1 m rises from 1.5% (F3Loc) to 5.6%, and success at 1 m from 22.4% to 55.5%; on the harder Gibson(g) set the method reaches 9.4% success at 0.1 m, above the 8.3% of the best prior representation-learning baseline. The gains come without any test-time depth, extra semantic labels, or added inference cost, because the priors are folded into the same observation encoder used by the existing pipeline.

What carries the argument

The load-bearing object is the visual encoder $F_\theta$, a UNet-style ResNet50, pretrained with two contrastive objectives. GCVI builds positive pairs from frame pairs whose pixels correspond to the same 3D points, established by projecting depth with the pinhole model and requiring the world coordinates to agree within 2 cm for at least 30% of pixels. VSAG builds positive pairs from an RGB frame and the cropped point-cloud chunk of the scene surface reconstruction falling in that frame's frustum, at 2 cm resolution, aligning color with geometry. Both use the PointInfoNCE loss, and the second objective also trains a PointNet++ branch $F_\phi$; after pretraining, $F_\theta$ is dropped into the F3Loc observation model and fine-tuned end-to-end with the histogram filter, so the geometric knowledge enters only through the encoder weights and adds no runtime cost.

What would settle it

Reproduce the same fine-tuning pipeline on Gibson(g) but initialize the observation encoder with a contrastive pretraining that uses the same ScanNet frames with the correspondence threshold relaxed from 2 cm to 20 cm, or with the VSAG alignment removed. If success at 0.1 m stays near the reported 9.4% rather than falling toward the 4.3% scratch baseline, the hard geometric constraints are not what drives the gain.

Watch

Extended reading notes

Core claim

The paper's central claim is that 3D geometric priors can be distilled into a 2D visual encoder, and that this distillation is what lifts floorplan localization accuracy. Two priors are defined. Geometry-Constrained View Invariance (GCVI) treats two frames as a positive pair when, via known poses and depth, at least 30% of their pixels project to 3D points within 2 cm of each other. View-Scene Aligned Geometry (VSAG) aligns each RGB frame with the point-cloud chunk of the scanned surface that lies in its view frustum, again under a 2 cm pixel-to-point correspondence. Both are trained with a PointInfoNCE contrastive loss on ScanNet, using BundleFusion poses and reconstructions, with no manual annotations. The pretrained encoder is then transferred into the observation model of F3Loc, a front-end ray and depth predictor plus back-end histogram filter, and fine-tuned on the target floorplan datasets. The measured consequence is that localization success at tight tolerances rises well above the F3Loc baseline and above prior 2D and 3D contrastive pretraining methods, without changing what is computed at inference time.

Load-bearing premise

The load-bearing premise is that geometric invariance and view-to-scene alignment learned on ScanNet RGB-D scans with BundleFusion poses and 2 cm pixel/point constraints transfer to monocular RGB inference on Structured3D and Gibson, where no depth is available at test time.

Editorial extensions

If this is right

  • On Structured3D full, single-frame success at 0.1 m improves from 1.5% (F3Loc_s) to 5.6%, and success at 1 m from 22.4% to 55.5%.
  • On Gibson(g), single-frame success at 0.1 m reaches 9.4% versus 8.3% for the strongest prior representation-learning baseline, and long-sequence success at 0.2 m reaches 70.3% for the Gibson(g)-trained single-frame model versus 35.1% for F3Loc_s.
  • Ablations show GCVI and VSAG contribute roughly equally, and either prior alone leaves success at 0.1 m near 6% on Gibson(g), well below the combined 9.4%.
  • Because the pretrained encoder replaces only the initialization of the same observation model, the improvements add no extra inference computation.
  • The pretraining needs no semantic labels: positive and negative pairs are defined purely by scan geometry, so the recipe extends to any RGB-D scan collection with poses.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the transfer story holds, the same self-supervised pretraining recipe could be dropped into other localization front-ends, such as particle-filter observation models, retrieval-based pose estimators, or map-matching networks; the paper demonstrates only the F3Loc case.
  • The 2 cm threshold and 30% overlap rule are design choices, not theorems. A natural stress test is to coarsen the threshold to, say, 20 cm and see whether Gibson(g) success at 0.1 m degrades; if it does not, the hard geometric alignment is not the active ingredient.
  • A practical corollary the paper does not test directly: since depth is used only in pretraining, the same encoder could be deployed on an ordinary phone camera with no depth sensor, as long as the scene floorplan and rough gravity direction are available.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. This paper proposes to improve 2D visual floorplan localization (FLoc) by pre-training the visual encoder with two self-supervised contrastive objectives on ScanNet RGB-D data: geometry-constrained view invariance (GCVI), which uses 2 cm 3D-point correspondences between frames as positive pairs, and view-scene aligned geometry (VSAG), which aligns RGB frames with surface point-cloud chunks. The pre-trained encoder is then fine-tuned inside the F3Loc observation model for single-frame, multi-frame, and adaptive inference. Experiments on Structured3D and Gibson report consistent improvements in success rate at several accuracy thresholds over F3Loc and other pretraining baselines, plus ablations and qualitative trajectory comparisons.

Significance. If the improvements are truly caused by the hard geometric constraints, the contribution is practically valuable: it is a simple, label-free encoder pre-training recipe that improves FLoc success and accuracy at no additional inference cost. The paper's main strengths are the clear formulation of the two priors, the use of public ScanNet data, and the breadth of the evaluation (two datasets, three inference modes, four metrics). The weakness is that the experimental design does not yet isolate the geometric prior from correlated factors such as indoor-domain pretraining data, the contrastive objective itself, and the fine-tuning protocol. The absence of variance estimates further limits the strength of the claims. With additional controlled experiments and uncertainty reporting, the paper could fully support its central claim.

major comments (3)
  1. [§4.3 (Tables 4 and 5)] The central claim that the hard 3D geometric constraints (GCVI, VSAG) drive the gains is confounded. Table 5 varies pretraining data, objective, and input modality simultaneously across baselines (SimCLR on ImageNet, CRL/Ego2-MAP/ECL on embodied RGB or semantic maps, SPA on neural rendering), while Table 4 only removes one of the two proposed losses and leaves the other in place; there is no control that keeps the same ScanNet RGB-D frame pairs and 840k pair budget while replacing the geometric correspondence loss with a non-geometric contrastive loss. Consequently, the improvement over F3Loc could be explained by domain-matched indoor self-supervised pretraining rather than by the 2 cm pixel/point constraints. Please add a control (e.g., the same ScanNet data and pair budget with a SimCLR-style or random-pair contrastive loss) and report it in Table 4 or Table 5.
  2. [All tables, especially Table 5] All success rates are single numbers from one run; no error bars, multiple seeds, or significance tests are reported. The smallest claimed advantage over the strongest baseline is +1.1 percentage points at SR@0.1m in Table 5 (Ours vs. SPA), which is within the range of typical run-to-run variance for this kind of fine-tuning. Please report mean and standard deviation over at least three seeds (or a paired significance test) for the main comparisons, and in particular for Table 5, so the reader can assess whether the remaining gains over SPA are reliable.
  3. [§4.1–4.2 (Tables 1–3)] The paper does not state whether the F3Loc baseline numbers are re-implemented under the same fine-tuning protocol (optimizer, epochs, loss, ray counts) or taken from the original F3Loc paper. Since the paper's headline gains are relative to F3Loc, this matters: if the baselines use a different training budget, part of the improvement may be due to fine-tuning configuration rather than to the pretrained geometric priors. Please clarify the source of each baseline number and, if available, provide a same-protocol re-implementation of F3Loc to make the comparison controlled.
minor comments (4)
  1. [§4.2] The sentence 'Ours_f further benefits from the performance gains of Ours_f and Ours_m' should read 'Ours_s and Ours_m'.
  2. [Eq. (3)] The denominator as printed is missing the summation symbol; the formula is ambiguous. Please use a clear summation notation, e.g., sum over all negative keys, to match the PointInfoNCE definition.
  3. [§3.2 and Abstract] The method requires RGB-D depth maps and camera poses (from SLAM) to define the geometric correspondences during pre-training; the statement 'no additional geometric annotations are required' should be qualified to clarify that metric depth and poses are used as supervision signals, though they are not manually annotated.
  4. [§4.3] The text says the contributions of GCVI and VSAG are 'almost equal'; Table 4 shows similar SR@1m values (38.3 vs. 37.5) but the difference at SR@0.1m is small (6.2 vs. 6.4). This summary is acceptable, but given the absence of variance estimates, the 'almost equal' claim should be stated more cautiously.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the claimed gains rest on held-out benchmark comparisons, not on equations that reduce to their inputs.

full rationale

The paper's central claim, that contrastive pre-training with GCVI and VSAG priors on ScanNet improves visual floorplan localization on Structured3D and Gibson, is supported by held-out test evaluations. Pre-training uses ScanNet RGB-D scans and BundleFusion poses, while all FLoc evaluations are on Structured3D and Gibson test splits, so the reported metrics are not the pre-training objectives or the pair-construction thresholds. The downstream fine-tuning loss (Eq. 5) supervises predicted 2D-ray depths against ground-truth depths, and the reported success rates come from comparing localized poses against ground-truth poses; none of these quantities is defined in terms of the reported outcome. Equations (1)-(3) and the PointInfoNCE losses define positive pairs geometrically, but those pairs are inputs to pre-training, not outputs of the evaluation. The only self-citations (Refs. [7] and [8]) appear as related-work context and as one baseline in Table 5; they do not justify the central transfer claim, and no uniqueness or ansatz is imported from them. The ablation in Table 4 is incomplete (no 'neither prior' row), and Table 5 varies pretraining data and objective simultaneously, which weakens causal attribution but is a confound, not circularity. Under the requirement to exhibit a specific reduction from output to input, no such step is present.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim is empirical: 3D-geometric contrastive pre-training improves floorplan localization. The load-bearing choices are the geometric correspondence thresholds and the transferability of the encoder from ScanNet to the evaluation datasets. No new physical entities are introduced.

free parameters (4)
  • 2 cm correspondence threshold = 2 cm
    Defines positive pixel/point pairs in GCVI and VSAG (Sec 3.2); chosen manually to balance accuracy and cost, and directly controls the hard geometric constraint credited for the performance gains.
  • 30% pixel correspondence ratio = 30%
    Minimum fraction of corresponding pixels for a positive image pair in GCVI (Sec 3.2); no sensitivity analysis is provided.
  • Contrastive temperature tau = 0.07
    Temperature coefficient in the PointInfoNCE losses (Sec 3.2 and Implementation Details); a standard value but chosen without reported ablation.
  • GCVI and VSAG loss weighting = 1:1
    Weighting of L_GCVI and L_VSAG during pre-training (Implementation Details); no sensitivity analysis is reported.
assumptions (4)
  • domain assumption ScanNet SLAM reconstructions and camera poses are accurate enough to support 2 cm-level correspondences.
    Sec 3.2 states that poses and 3D reconstructions are obtained automatically using BundleFusion [14] without manual verification; any drift would create false positives or negatives in contrastive learning.
  • domain assumption At test time, camera intrinsics, gravity direction, and relative poses between frames are known.
    Stated in Sec 3.1; this is inherited from F3Loc and limits deployment scenarios, but the ray-based observation model requires these inputs.
  • domain assumption The ScanNet-pretrained visual encoder transfers to Structured3D and Gibson RGB-only FLoc without test-time depth.
    This is the load-bearing transfer premise of Sec 3.4; the only evidence is the benchmark tables, not a formal argument or released pretrained weights.
  • standard math Pinhole camera projection and rigid transforms in Eqs. (1)-(2) correctly map pixels to world coordinates.
    Standard imaging geometry used to establish all correspondences; treated as background in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Perspective from a Higher Dimension: Can 3D Geometric Priors Help Visual Floorplan Localization?." pith.science (2026). https://pith.science/paper/YC66K7KC

@misc{pith2026250718881,
  author       = {Pith},
  title        = {Pith review of: Perspective from a Higher Dimension: Can 3D Geometric Priors Help Visual Floorplan Localization?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YC66K7KC}},
  note         = {Machine review of arXiv:2507.18881}
}
read the original abstract

Since a building's floorplans are easily accessible, consistent over time, and inherently robust to changes in visual appearance, self-localization within the floorplan has attracted researchers' interest. However, since floorplans are minimalist representations of a building's structure, modal and geometric differences between visual perceptions and floorplans pose challenges to this task. While existing methods cleverly utilize 2D geometric features and pose filters to achieve promising performance, they fail to address the localization errors caused by frequent visual changes and view occlusions due to variously shaped 3D objects. To tackle these issues, this paper views the 2D Floorplan Localization (FLoc) problem from a higher dimension by injecting 3D geometric priors into the visual FLoc algorithm. For the 3D geometric prior modeling, we first model geometrically aware view invariance using multi-view constraints, i.e., leveraging imaging geometric principles to provide matching constraints between multiple images that see the same points. Then, we further model the view-scene aligned geometric priors, enhancing the cross-modal geometry-color correspondences by associating the scene's surface reconstruction with the RGB frames of the sequence. Both 3D priors are modeled through self-supervised contrastive learning, thus no additional geometric or semantic annotations are required. These 3D priors summarized in extensive realistic scenes bridge the modal gap while improving localization success without increasing the computational burden on the FLoc algorithm. Sufficient comparative studies demonstrate that our method significantly outperforms state-of-the-art methods and substantially boosts the FLoc accuracy. All data and code will be released after the anonymous review.

Figures

Figures reproduced from arXiv: 2507.18881 by the authors.

Figure 1
Figure 1. (a) An illustration of the visual Floorplan Localization (FLoc) task. (b) Our 3D geometric prior-enhanced visual FLoc [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Our contrastive pre-training includes the modelings of GCVI ( [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 2
Figure 2. PointInfoNCE is likewise used as the contrastive loss [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: Comparison of the localization success rate using [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]
Figure 4
Figure 4. Figure 4: Qualitative comparative studies in a scene without [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: A qualitative comparative study in a larger-scale scene with complex furniture. Our single-frame localization is highly [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

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

    cs.RO 2026-07 conditional novelty 6.0 of 10

    CF2Loc replaces ray-matching in visual floorplan localization with a diffusion-based multimodal pose sampler plus a crop-based residual refiner, reporting state-of-the-art recall on S3D (full) and ZInD.

Reference graph

Works this paper leans on

69 extracted references · 49 canonical work pages · cited by 1 Pith paper

  1. [1]

    Relja Arandjelovic, Petr Gronat, Akihiko Torii, Tomas Pajdla, and Josef Sivic

  2. [2]

    Pattaramanee Arsomngern, Sarana Nutanong, and Supasorn Suwajanakorn. 2023. Learning Geometric-Aware Properties in 2D Representation Using Lightweight CAD Models, or Zero Real 3D Pairs. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 21371–21381

  3. [3]

    Vassileios Balntas, Shuda Li, and Victor Prisacariu. 2018. Relocnet: Continuous metric learning relocalisation using neural nets. InProceedings of the European conference on computer vision (ECCV). 751–767

  4. [4]

    Gary Bishop, Greg Welch, et al. 2001. An introduction to the kalman filter.Proc of SIGGRAPH, Course8, 27599-23175 (2001), 41

  5. [5]

    Federico Boniardi, Abhinav Valada, Rohit Mohan, Tim Caselitz, and Wolfram Burgard. 2019. Robot localization in floor plans using a room layout edge extrac- tion network. In2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 5291–5297

  6. [6]

    Eric Brachmann, Alexander Krull, Sebastian Nowozin, Jamie Shotton, Frank Michel, Stefan Gumhold, and Carsten Rother. 2017. Dsac-differentiable ransac for camera localization. InProceedings of the IEEE conference on computer vision and pattern recognition. 6684–6692

  7. [7]

    Bolei Chen, Jiaxu Kang, Ping Zhong, Yongzheng Cui, Siyi Lu, Yixiong Liang, and Jianxin Wang. 2023. Think holistically, act down-to-earth: A semantic navigation strategy with continuous environmental representation and multi-step forward planning.IEEE Transactions on Circuits and Systems for Video Technology(2023)

  8. [8]

    Bolei Chen, Jiaxu Kang, Ping Zhong, Yixiong Liang, Yu Sheng, and Jianxin Wang

Show all 69 references
  1. [9]

    Changan Chen, Rui Wang, Christoph Vogel, and Marc Pollefeys. 2024. F3Loc: Fusion and Filtering for Floorplan Localization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 18029–18038

  2. [10]

    Nenglun Chen, Lei Chu, Hao Pan, Yan Lu, and Wenping Wang. 2022. Self- supervised image representation learning with geometric set consistency. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 19292–19302

  3. [11]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. InInterna- tional conference on machine learning. PMLR, 1597–1607

  4. [12]

    Hang Chu, Dong Ki Kim, and Tsuhan Chen. 2015. You are here: Mimicking the human thinking process in reading floor-plans. InProceedings of the IEEE International Conference on Computer Vision. 2210–2218

  5. [13]

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner. 2017. Scannet: Richly-annotated 3d reconstructions of indoor scenes. InProceedings of the IEEE conference on computer vision and pattern recognition. 5828–5839

  6. [14]

    Angela Dai, Matthias Nießner, Michael Zollhöfer, Shahram Izadi, and Christian Theobalt. 2017. Bundlefusion: Real-time globally consistent 3d reconstruction using on-the-fly surface reintegration.ACM Transactions on Graphics (ToG)36, 4 (2017), 1

  7. [15]

    Frank Dellaert, Dieter Fox, Wolfram Burgard, and Sebastian Thrun. 1999. Monte carlo localization for mobile robots. InProceedings 1999 IEEE international confer- ence on robotics and automation (Cat. No. 99CH36288C), Vol. 2. IEEE, 1322–1328

  8. [16]

    Mingyu Ding, Zhe Wang, Jiankai Sun, Jianping Shi, and Ping Luo. 2019. CamNet: Coarse-to-fine retrieval for camera re-localization. InProceedings of the IEEE/CVF International Conference on Computer Vision. 2871–2880

  9. [17]

    Yilun Du, Chuang Gan, and Phillip Isola. 2021. Curious representation learning for embodied intelligence. InProceedings of the IEEE/CVF International Conference on Computer Vision. 10408–10417

  10. [18]

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. 2020. Mo- mentum contrast for unsupervised visual representation learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 9729–9738

  11. [19]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. InProceedings of the IEEE conference on computer vision and pattern recognition. 770–778

  12. [20]

    Yicong Hong, Yang Zhou, Ruiyi Zhang, Franck Dernoncourt, Trung Bui, Stephen Gould, and Hao Tan. 2023. Learning navigational visual representations with semantic map supervision. InProceedings of the IEEE/CVF International Conference on Computer Vision. 3055–3067

  13. [21]

    Henry Howard-Jenkins and Victor Adrian Prisacariu. 2022. LaLaLoc++: Global floor plan comprehension for layout localisation in unvisited environments. In European Conference on Computer Vision. Springer, 693–709

  14. [22]

    Henry Howard-Jenkins, Jose-Raul Ruiz-Sarmiento, and Victor Adrian Prisacariu

  15. [23]

    Sergey Ioffe. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift.arXiv preprint arXiv:1502.03167(2015)

  16. [24]

    Seigo Ito, Felix Endres, Markus Kuderer, Gian Diego Tipaldi, Cyrill Stachniss, and Wolfram Burgard. 2014. W-rgb-d: floor-plan-based indoor global localization using a depth camera and wifi. In2014 IEEE international conference on robotics and automation (ICRA). IEEE, 417–422

  17. [25]

    Rico Jonschkowski and Oliver Brock. 2016. End-to-end learnable histogram filters. (2016)

  18. [26]

    Peter Karkus, David Hsu, and Wee Sun Lee. 2018. Particle filter networks with application to visual localization. InConference on robot learning. PMLR, 169–178

  19. [27]

    Alex Kendall, Matthew Grimes, and Roberto Cipolla. 2015. Posenet: A convolu- tional network for real-time 6-dof camera relocalization. InProceedings of the IEEE international conference on computer vision. 2938–2946

  20. [28]

    Diederik P Kingma. 2014. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980(2014)

  21. [29]

    Obin Kwon, Jeongho Park, and Songhwai Oh. 2023. Renderable neural radiance map for visual navigation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 9099–9108

  22. [30]

    Jiaxin Li, Weiqi Huang, Zan Wang, Wei Liang, Huijun Di, and Feng Liu

  23. [31]

    Xinran Liang, Anthony Han, Wilson Yan, Aditi Raghunathan, and Pieter Abbeel

  24. [32]

    Liu Liu, Hongdong Li, and Yuchao Dai. 2017. Efficient global 2d-3d matching for camera localization in a large-scale 3d map. InProceedings of the IEEE International Conference on Computer Vision. 2372–2381

  25. [33]

    Qiming Liu, Haoran Xin, Zhe Liu, and Hesheng Wang. 2024. Integrating neural radiance fields end-to-end for cognitive visuomotor navigation.IEEE Transactions on Pattern Analysis and Machine Intelligence(2024)

  26. [34]

    FloNa: Floor Plan Guided Embodied Visual Navigation.arXiv preprint arXiv:2412.18335(2024)

  27. [35]

    Oscar Mendez, Simon Hadfield, Nicolas Pugeault, and Richard Bowden. 2020. SeDAR: reading floorplans like a human—using deep learning to enable human- inspired localisation.International Journal of Computer Vision128, 5 (2020), 1286–1310

  28. [36]

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. 2021. Nerf: Representing scenes as neural radiance fields for view synthesis.Commun. ACM65, 1 (2021), 99–106

  29. [37]

    Zhixiang Min, Naji Khosravan, Zachary Bessinger, Manjunath Narayana, Sing Bing Kang, Enrique Dunn, and Ivaylo Boyadzhiev. 2022. Laser: Latent space rendering for 2d visual localization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 11122–11131

  30. [38]

    Matthias Nießner, Michael Zollhöfer, Shahram Izadi, and Marc Stamminger. 2013. Real-time 3D reconstruction at scale using voxel hashing.ACM Transactions on Graphics (ToG)32, 6 (2013), 1–11

  31. [39]

    Oscar Mendez, Simon Hadfield, Nicolas Pugeault, and Richard Bowden. 2018. Sedar-semantic detection and ranging: Humans can localise without lidar, can robots?. In2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 6053–6060

  32. [40]

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. 2015. U-net: Convolu- tional networks for biomedical image segmentation. InMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings...

  33. [41]

    Paul-Edouard Sarlin, Cesar Cadena, Roland Siegwart, and Marcin Dymczyk

  34. [42]

    Paul-Edouard Sarlin, Daniel DeTone, Tsun-Yi Yang, Armen Avetisyan, Julian Straub, Tomasz Malisiewicz, Samuel Rota Bulo, Richard Newcombe, Peter Kontschieder, and Vasileios Balntas. 2023. Orienternet: Visual localization in 2d public maps with neural matching. InProceedings of ...

  35. [43]

    Paul-Edouard Sarlin, Mihai Dusmanu, Johannes L Schönberger, Pablo Speciale, Lukas Gruber, Viktor Larsson, Ondrej Miksik, and Marc Pollefeys. 2022. Lamar: Benchmarking localization and mapping for augmented reality. InEuropean MM ’25, October 27–31, 2025, Dublin, Ireland. Bolei...

  36. [44]

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. 2017. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems30 (2017)

  37. [45]

    Torsten Sattler, Bastian Leibe, and Leif Kobbelt. 2016. Efficient & effective pri- oritized matching for large-scale image-based localization.IEEE transactions on pattern analysis and machine intelligence39, 9 (2016), 1744–1756

  38. [46]

    Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. 2021. LoFTR: Detector-free local feature matching with transformers. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 8922–8931

  39. [47]

    Julien Valentin, Matthias Nießner, Jamie Shotton, Andrew Fitzgibbon, Shahram Izadi, and Philip HS Torr. 2015. Exploiting uncertainty in regression forests for accurate camera relocalization. InProceedings of the IEEE conference on computer vision and pattern recognition. 4400–4408

  40. [48]

    Rudolph Van Der Merwe, Arnaud Doucet, Nando De Freitas, and Eric Wan. 2000. The unscented particle filter.Advances in neural information processing systems 13 (2000)

  41. [49]

    A Vaswani. 2017. Attention is all you need.Advances in Neural Information Processing Systems(2017)

  42. [50]

    Paul-Edouard Sarlin, Ajaykumar Unagar, Mans Larsson, Hugo Germain, Carl Toft, Viktor Larsson, Marc Pollefeys, Vincent Lepetit, Lars Hammarstrand, Fredrik Kahl, et al. 2021. Back to the feature: Learning robust camera localization from pixels to pose. InProceedings of the IEEE/...

  43. [51]

    Zihan Wang, Xiangyang Li, Jiahao Yang, Yeqi Liu, Junjie Hu, Ming Jiang, and Shuqiang Jiang. 2024. Lookahead Exploration with Neural Radiance Representa- tion for Continuous Vision-Language Navigation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

  44. [52]

    Wera Winterhalter, Freya Fleckenstein, Bastian Steder, Luciano Spinello, and Wolfram Burgard. 2015. Accurate indoor localization for RGB-D smartphones and tablets given 2D floor plans. In2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 3138–3143

  45. [53]

    Jian Wu, Liwei Ma, and Xiaolin Hu. 2017. Delving deeper into convolutional neural networks for camera relocalization. In2017 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 5644–5651

  46. [54]

    Fei Xia, Amir R Zamir, Zhiyang He, Alexander Sax, Jitendra Malik, and Silvio Savarese. 2018. Gibson env: Real-world perception for embodied agents. In Proceedings of the IEEE conference on computer vision and pattern recognition. 9068–9079

  47. [55]

    Saining Xie, Jiatao Gu, Demi Guo, Charles R Qi, Leonidas Guibas, and Or Litany

  48. [56]

    Florian Walch, Caner Hazirbas, Laura Leal-Taixe, Torsten Sattler, Sebastian Hilsen- beck, and Daniel Cremers. 2017. Image-based localization using lstms for struc- tured feature correlation. InProceedings of the IEEE international conference on computer vision. 627–637

  49. [57]

    Luwei Yang, Ziqian Bai, Chengzhou Tang, Honghua Li, Yasutaka Furukawa, and Ping Tan. 2019. Sanet: Scene agnostic network for camera localization. In Proceedings of the IEEE/CVF international conference on computer vision. 42–51

  50. [58]

    Yuqi Zhang, Han Luo, and Yinjie Lei. 2024. Towards CLIP-driven Language-free 3D Visual Grounding via 2D-3D Relational Enhancement and Consistency. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 13063–13072

  51. [59]

    Jia Zheng, Junfei Zhang, Jing Li, Rui Tang, Shenghua Gao, and Zihan Zhou

  52. [60]

    Qunjie Zhou, Torsten Sattler, and Laura Leal-Taixe. 2021. Patch2pix: Epipolar- guided pixel-level correspondences. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 4669–4678

  53. [61]

    Haoyi Zhu, Honghui Yang, Yating Wang, Jiange Yang, Limin Wang, and Tong He. 2024. Spa: 3d spatial-awareness enables effective embodied representation. arXiv preprint arXiv:2410.08208(2024)

  54. [63]

    Karmesh Yadav, Ram Ramrakhya, Arjun Majumdar, Vincent-Pierre Berges, Sachit Kuhar, Dhruv Batra, Alexei Baevski, and Oleksandr Maksymets. 2023. Offline visual representation learning for embodied navigation. InWorkshop on Reincar- nating Reinforcement Learning at ICLR 2023

  55. [67]

    InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IX 16

    Structured3d: A large photo-realistic dataset for structured 3d modeling. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IX 16. Springer, 519–535

  56. [2017]

    NetVLAD: CNN architecture for weakly supervised place recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence(2017), 1–1

  57. [2019]

    In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    From coarse to fine: Robust hierarchical localization at large scale. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 12716–12725

  58. [2020]

    InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part III 16

    Pointcontrast: Unsupervised pre-training for 3d point cloud understanding. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part III 16. Springer, 574–591

  59. [2021]

    In Proceedings of the IEEE/CVF International Conference on Computer Vision

    Lalaloc: Latent layout localisation in dynamic, unvisited environments. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 10107– 10116

  60. [2023]

    ALP: Action-Aware Embodied Learning for Perception.arXiv preprint arXiv:2306.10190(2023)

  61. [2024]

    InProceedings of the 32nd ACM Interna- tional Conference on Multimedia

    Embodied Contrastive Learning with Geometric Consistency and Behav- ioral Awareness for Object Navigation. InProceedings of the 32nd ACM Interna- tional Conference on Multimedia. 4776–4785

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.