Pith. sign in

REVIEW 3 major objections 5 minor 41 references

Cascaded Parallel Filtering for Memory-Efficient Image-Based Localization

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Compact binary descriptors, filtered in a cascaded parallel pipeline, match full-SIFT localization accuracy at about a fifth of the memory.

desk verdict Solid engineering paper on memory-efficient localization with binary descriptors; the per-dataset tuning and missing recall analysis are the real caveats, not deal-breakers. read the letter →

arxiv 1908.06141 v1 pith:OWFLULSB submitted 2019-08-16 cs.CV

Wentao Cheng , Weisi Lin , Kan Chen , Xinfeng Zhang This is my paper · ORCID
classification cs.CV
keywords image-basedlocalization6DOFcameraposeestimationHammingembeddingbinaryfeaturesmatchfilteringstructure-from-motioncascadedfilters
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

The paper claims that image-based localization does not need a full SIFT descriptor per 3D point to be accurate. It proposes a pipeline in which each 3D point is stored as a short binary signature via Hamming Embedding, and wrong candidate 2D-3D matches are removed in three successive filter stages: feature-level, visibility-level, and geometry-level. At each of the first two stages the pipeline runs two parallel tasks with different strictness, one preserving potentially correct matches and one producing high-confidence matches for the next stage. The result, the authors argue, is localization accuracy competitive with state-of-the-art SIFT-based methods on Dubrovnik, RobotCar Seasons, and Aachen Day-Night, and competitive on SF-0, while storing roughly five times less descriptor memory than Active Search. They also add a quality-aware spatial reconfiguration step that spreads matches across the image before RANSAC, and a principal focal length step that picks the auxiliary pose by median focal length among top hypotheses.

What carries the argument

The central machinery is a cascade of three filter stages over Hamming-Embedding matches, with parallel strict and relaxed branches in the first two stages. Feature-wise filtering uses the bilateral Hamming ratio test, which scores a match by comparing its Hamming distance against the average distance within both the query image's feature space and the 3D model's feature space; matches survive to a relaxed match pool if their score is positive, and high-scoring matches become Feature-wisely Confident. Visibility-wise filtering votes database images using only confident matches on the bipartite visibility graph, retrieves top-k and top-k1 ranked images, and uses a two-step selection to infer potentially correct matches from co-visibility with confident matches. Geometry-wise filtering computes an auxiliary pose from spatially reconfigured matches and uses reprojection error to filter the retained pool, then a principal-focal-length selection step and a P3P solver produce the final pose.

What would settle it

Using a benchmark with ground-truth 2D-3D correspondences from SfM triangulation, measure the recall of true correspondences immediately after the initial visual-word assignment and tau-threshold step. If that recall is below the final localization recall on the same data, the cascade cannot be the only source of accuracy; if it is far above, the later filters are doing the real work. Either way, the measurement locates where matches are lost.

Watch

Extended reading notes

Core claim

The central claim is that the ambiguity introduced by compact binary descriptors can be absorbed by a cascade of filters, so a structure-based localization system can be memory-efficient without sacrificing accuracy. The paper's specific assertion is that a Hamming Embedding representation with a bilateral Hamming ratio test, visibility voting over a bipartite point-image graph, and a final geometry filter using an auxiliary pose computed from high-confidence matches lets a 64-bit or 128-bit signature per visual word match or beat SIFT-based methods on standard benchmark datasets. On Dubrovnik the median error drops to 0.64 m with 794 of 800 queries localized; on Aachen Day-Night it reaches 76.7% of day queries and 25.5% of night queries within the high-precision interval, and on RobotCar Seasons it is second only to a semantic method that uses extra labelled training data. The memory comparison against Active Search is 0.14 GB versus 0.75 GB on Dubrovnik, 0.52 GB versus 2.72 GB on RobotCar, and 0.14 GB versus 0.76 GB on Aachen.

Load-bearing premise

The whole pipeline can only recover matches that survive the very first assignment: a true query-to-3D correspondence must fall into the same visual word and have binary Hamming distance at most tau, and if the compact signature pushes it past that gate, no visibility or geometry filter can ever see it.

Editorial extensions

If this is right

  • Memory-efficient localization on mobile or embedded devices becomes plausible, since storing 8-byte signatures instead of 128-byte descriptors per visual word sharply cuts per-point storage.
  • The relaxed-plus-strict parallel branching offers a general recipe: keep a permissive match pool for recall while a strict branch supplies seeds for progressively stronger filters.
  • On the tested benchmarks, compact binary descriptors can be used without the accuracy penalty previously expected, suggesting descriptor compactness and match disambiguation are separable problems.
  • Selecting the auxiliary pose by the median focal length among near-best RANSAC hypotheses can beat selecting the hypothesis with the most inliers when correct matches are sparse.
  • The strong Aachen night-time results indicate that a strong filter cascade can handle severe appearance change even without semantic priors or known gravity direction.

Reading between the lines

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

  • A natural extension would be to replace the hand-set thresholds (tau, alpha, k, k1, N, theta) with learned or per-query adaptive values; the paper's own ablation shows sensitivity to these thresholds.
  • The visibility and geometry stages are agnostic to the underlying descriptor, so the cascade should compose with learned binary descriptors from a CNN; replacing SIFT extraction with a learned detector is the next obvious experiment.
  • The two-branch design could also run in reverse for hierarchical localization: a cheap strict branch produces a coarse pose, which then verifies a large relaxed match pool, potentially reducing retrieval computation.
  • Because the SF-0 results are weakest for coarse-level 5-30 m localization, a practical system would fuse GPS or retrieval priors when available; the paper notes that GPS helps but does not explore the fusion.
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 / 5 minor

Summary. This paper proposes a memory-efficient structure-based image localization pipeline based on binary Hamming Embedding. The method stores a compact binary signature per visual word for each 3D point, establishes 2D-3D matches by visual-word assignment and a Hamming distance threshold, and then filters these matches in a cascade of feature-, visibility-, and geometry-wise steps. Two parallel criteria are used at the feature and visibility stages: a relaxed path preserves a large match pool, while a strict path extracts high-confidence matches that support an auxiliary camera pose. The auxiliary pose is obtained using a quality-aware spatial reconfiguration and a principal-focal-length selection heuristic, and is then used to filter the relaxed pool by reprojection error. The authors evaluate on Dubrovnik, RobotCar Seasons, Aachen Day-Night, and SF-0, and report pose accuracy competitive with full-descriptor methods while using roughly five times less descriptor memory than Active Search.

Significance. If the reported results are robust, the paper makes a useful contribution to memory-efficient localization: it replaces 128-byte integer SIFT mean descriptors with 64-bit binary signatures while claiming accuracy competitive with full-descriptor methods on four established benchmarks. The cascaded parallel filter design is interesting, and the ablation on Dubrovnik (Table 6) indicates that each major component contributes to the final accuracy. The source code is released, which supports reproducibility. However, the load-bearing evidence is incomplete in two ways: the recall of the initial feature-wise matching stage is never measured, and several thresholds are tuned per dataset without a sensitivity analysis, so the reported gains could be partly determined by the chosen parameter values. The memory comparison is also narrower than the 'state-of-the-art' wording suggests.

major comments (3)
  1. [Section 2.2, Eq. (8)] Equation (8) defines the updated match score as E'(m) = E(m) + sum_{d in d(k)} (alpha/2) ln(1 + omega^d_VFC / omega^d_VNFC). As written, the summation runs over all top-k database images and the added term does not depend on m at all: the same increment is added to every VNFC match of a given query. This is inconsistent with the surrounding explanation, which says that a VNFC match should be promoted only if its 3D point is co-visible with VFC matches in the top-ranked database images. The formula should presumably sum only over database images that observe the 3D point of m, or should otherwise involve the point-image incidence. Please correct the equation and the corresponding description in Algorithm 1; as printed, the two-step match selection is not reproducible and the reported benefit of VFC-I matches is not traceable to the stated mechanism.
  2. [Section 2.1 and Eq. (5)] In the feature-wise stage, a 2D-3D match is created only if the query descriptor and a 3D point's integer mean descriptor are assigned to the same visual word and h(s_q, s_p) <= tau. All later steps filter within the resulting pool M (Eq. 5), so any true correspondence that fails this condition is permanently lost. The paper tunes tau to 19, 16, and 32 on Dubrovnik/RobotCar, Aachen, and SF-0, respectively (Section 3.2), but reports neither the recall of this stage nor the sensitivity of the final pose accuracy to tau. If the true-match pass rate is low, the reported localization accuracy would be bounded by the binary quantization and threshold choice rather than by the cascaded filtering design, and the claimed generality to unseen datasets would not be supported. Please provide, for at least one benchmark (e.g., Dubrovnik), a curve of feature-stage recall versus tau computed from ground-truth 2D-3D correspondences, and the corresponding final localization accuracy versus tau.
  3. [Section 3.2] The parameters tau, alpha, B, k, k1, N, beta, and theta are assigned separately per dataset (e.g., tau=19 for Dubrovnik/RobotCar, tau=16 for Aachen, B=128 with tau=32 and alpha=0.4 for SF-0), and the paper does not describe a validation procedure or a selection rule that would apply to a new dataset. This raises the risk that the reported numbers are partly the result of benchmark-specific tuning. Please add a sensitivity analysis on at least one dataset, varying the most influential thresholds (e.g., tau and alpha on Dubrovnik), and state how the values would be chosen in a deployment scenario without ground-truth query poses.
minor comments (5)
  1. [Table 5] The memory comparison is limited to Active Search among the structure-based methods compared in Tables 3 and 4 and Figure 3. The text justifies this by calling AS the most memory-efficient state-of-the-art structure-based method, but no memory figures are given for the other compared methods, so the broader 'state-of-the-art' memory claim cannot be checked from the paper. Please either report the descriptor/vocabulary memory of the other methods or explicitly scope the claim to the comparison with AS.
  2. [Section 3.2] The sentence 'For a fair comparison on the RobotCar Seasons and Aachen Day-Night dataset, we use a 3-point pose solver to compute the auxiliary camera pose' appears to contradict Section 2.3, which says the auxiliary camera pose is computed with a 4-point pose solver (P4P) when the focal length is unknown, and that the final pose uses P3P. Please clarify which solver is used for the auxiliary pose on these datasets.
  3. [Table 4] For the Aachen Day-Night dataset, the SMC* row is shown as '- -' without explanation; please state why SMC results are not reported for Aachen so that the comparison table is complete.
  4. [Section 3.4] There is a duplicated article in the text: 'heavily related to the the bilateral Hamming ratio test' should read 'heavily related to the bilateral Hamming ratio test.'
  5. [Algorithm 1] The return statement uses the notation M^d(k)_VFC and M^d(k)_VFC-I, but the subscript d(k) is not explicitly defined in the algorithm; please define it in the pseudocode or in the caption for readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported localization accuracy is measured against external benchmarks, and the filtering equations do not encode the output by construction.

full rationale

The paper's central claims are empirical and are tested on external benchmark datasets (Dubrovnik, RobotCar Seasons, Aachen Day-Night, SF-0) with ground-truth camera poses. The compact feature representation is inherited from prior external work on Hamming Embedding [15], and the Gaussian weighting function is adapted from [16]; neither is a self-citation. The proposed equations (1)-(10) define filtering and scoring rules, but they do not by themselves determine the reported localization accuracy: success is decided by RANSAC-based pose estimation against ground-truth poses, so the output numbers are not equivalent to the input definitions. The per-dataset choices of tau, alpha, B, k, and k1 in Section 3.2 are hyperparameter settings, not fitted quantities that are later renamed as predictions; the paper does not present these choices as derived results. The single self-citation ([8], a prior point-cloud simplification paper by overlapping authors) appears only in the related-work survey and is not load-bearing for the proposed method. The reader-identified concern that true matches outside the same visual word or beyond threshold tau are unrecoverable (Section 2.1) is a genuine recall limitation and a correctness risk, but it is not circular: it does not make the derived accuracy equal to an input. Overall, the derivation chain is self-contained against external benchmarks, so no circular step is present.

Assumptions & free parameters 11 free parameters · 6 assumptions · 0 invented entities

The system rests on standard structure-based localization assumptions: descriptors quantify visual similarity, visible 3D points in similar images cluster, co-visibility indicates correctness, and RANSAC pose hypotheses are reliable. There are no physically invented entities. The main burden is the large set of manually set hyperparameters, several of which are tuned per dataset in Section 3.2, so part of the reported performance is selected rather than predicted.

free parameters (11)
  • Hamming distance threshold tau = 19 (Dubrovnik, RobotCar), 16 (Aachen), 32 (SF-0)
    Hard threshold defining which 2D-3D matches are generated and retained in the feature-wise pool; tuned per dataset in Section 3.2.
  • Image-side Hamming ratio threshold phi = 0.3
    Threshold below which a match is rejected as ambiguous; Section 2.1 states it 'works well in practice' with no sensitivity analysis.
  • Feature confidence threshold alpha = 0.8 (medium-scale), 0.4 (SF-0)
    Separates feature-wisely confident matches from the rest of the match pool; tuned per dataset in Section 3.2.
  • Top-k database images for strict selection = 20
    Number of top-ranked database images used to select visibility-wisely confident matches; fixed in Section 3.2.
  • Top-k1 database images for relaxed pool = 100 (Dubrovnik, RobotCar), 50 (Aachen)
    Size of the visibility-wise match pool used for later geometry filtering; tuned per dataset in Section 3.2.
  • Maximum selected matches N for auxiliary pose = 100
    Overall quota for spatially reconfigured matches used in auxiliary camera pose estimation; set in Section 2.3.
  • VFC-I cap ratio beta = 0.33
    Limits the number of inferred VFC-I matches to beta times the number of VFC matches; set in Section 2.3.
  • Reprojection error threshold theta = 10 pixels
    Geometry-wise filtering threshold for accepting matches under the auxiliary camera pose; chosen in Section 2.3.
  • Binary signature length B = 64 (medium-scale), 128 (SF-0)
    Number of bits per Hamming Embedding signature; controls memory and distinctiveness, adjusted for SF-0 in Section 3.2.
  • RANSAC iterations = 1000 for auxiliary and final pose
    Computational budget for pose estimation; fixed in Section 3.2.
  • Visual vocabulary size = 10k (medium), 50k (SF-0)
    Number of visual words used for descriptor quantization; a general or specific vocabulary is trained per dataset in Section 3.2.
assumptions (6)
  • domain assumption Feature quantization: a 2D-3D match is generated only if the query descriptor and the 3D point's mean descriptor fall in the same visual word and have Hamming distance at most tau.
    Invoked in Section 2.1 during data pre-processing; if quantization misses true correspondences, no later filtering stage can recover them.
  • domain assumption Hamming distance after binary embedding is a valid proxy for descriptor distinctiveness.
    Used in Equations 1-5 for all feature scoring; the method assumes compact binary signatures preserve enough discrimination.
  • domain assumption Correct matches cluster in database images relevant to the query, and image voting with confident matches can identify those images.
    The visibility-wise filtering in Section 2.2 relies on top-k database image voting to select correct matches.
  • domain assumption Co-visibility with VFC matches in top database images indicates that a lower-confidence VNFC match is likely correct.
    Used in Equation 8 and Algorithm 1 to infer VFC-I matches from point-point visibility relationships.
  • ad hoc to paper The RANSAC hypothesis whose focal length is the median of the top-10 hypotheses is more reliable than the max-inlier hypothesis.
    This principal focal length heuristic is introduced in Section 2.3 without a formal proof, and the paper shows empirically that it improves accuracy.
  • domain assumption Benchmark ground truth poses and inlier thresholds are accurate enough to validate the method.
    All quantitative claims in Tables 3-5 and Figure 3 depend on the correctness of the Dubrovnik, RobotCar Seasons, Aachen Day-Night, and SF-0 ground truth.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cascaded Parallel Filtering for Memory-Efficient Image-Based Localization." pith.science (2026). https://pith.science/paper/OWFLULSB

@misc{pith2026190806141,
  author       = {Pith},
  title        = {Pith review of: Cascaded Parallel Filtering for Memory-Efficient Image-Based Localization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OWFLULSB}},
  note         = {Machine review of arXiv:1908.06141}
}
read the original abstract

Image-based localization (IBL) aims to estimate the 6DOF camera pose for a given query image. The camera pose can be computed from 2D-3D matches between a query image and Structure-from-Motion (SfM) models. Despite recent advances in IBL, it remains difficult to simultaneously resolve the memory consumption and match ambiguity problems of large SfM models. In this work, we propose a cascaded parallel filtering method that leverages the feature, visibility and geometry information to filter wrong matches under binary feature representation. The core idea is that we divide the challenging filtering task into two parallel tasks before deriving an auxiliary camera pose for final filtering. One task focuses on preserving potentially correct matches, while another focuses on obtaining high quality matches to facilitate subsequent more powerful filtering. Moreover, our proposed method improves the localization accuracy by introducing a quality-aware spatial reconfiguration method and a principal focal length enhanced pose estimation method. Experimental results on real-world datasets demonstrate that our method achieves very competitive localization performances in a memory-efficient manner.

Figures

Figures reproduced from arXiv: 1908.06141 by the authors.

Figure 1
Figure 1. Overview of the localization pipeline using our cascaded parallel filtering method. A cascade of feature-, visibility [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The influence of a uniform spatial distribution [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. The experimental results on the SF-0 dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Netvlad: Cnn architecture for weakly supervised place recognition

    Relja Arandjelovi ´c, Petr Gronat, Akihiko Torii, Tomas Pajdla, and Josef Sivic. Netvlad: Cnn architecture for weakly supervised place recognition. In Proc. CVPR,

    work page

  2. [2]

    All about vlad

    Relja Arandjelovi ´c and Andrew Zisserman. All about vlad. In Proc. CVPR, 2013. 1, 7

    work page 2013

  3. [3]

    Disloca- tion: Scalable descriptor distinctiveness for location recognition

    Relja Arandjelovi ´c and Andrew Zisserman. Disloca- tion: Scalable descriptor distinctiveness for location recognition. In Proc. ACCV, 2014. 3, 4

    work page 2014

  4. [4]

    Learning less is more-6d camera localization via 3d surface regression

    Eric Brachmann and Carsten Rother. Learning less is more-6d camera localization via 3d surface regression. In Proc. CVPR, 2018. 2

    work page 2018

  5. [5]

    A general solution to the p4p problem for camera with unknown focal length

    Martin Bujnak, Zuzana Kukelova, and Tomas Pajdla. A general solution to the p4p problem for camera with unknown focal length. In Proc. CVPR, 2008. 1, 6

    work page 2008

  6. [6]

    Toroidal con- straints for two-point localization under high outlier ratios

    Federico Camposeco, Torsten Sattler, Andrea Cohen, Andreas Geiger, and Marc Pollefeys. Toroidal con- straints for two-point localization under high outlier ratios. In Proc. CVPR, 2017. 1, 2, 6, 7

    work page 2017

  7. [7]

    Minimal scene descrip- tions from structure from motion models

    Song Cao and Noah Snavely. Minimal scene descrip- tions from structure from motion models. In Proc. CVPR, 2014. 2

    work page 2014

  8. [8]

    A data-driven point cloud simplification framework for city-scale image- based localization

    Wentao Cheng, Weisi Lin, Xinfeng Zhang, Michael Goesele, and Ming-Ting Sun. A data-driven point cloud simplification framework for city-scale image- based localization. IEEE Transactions on Image Pro- cessing, 26(1):262–275, 2017. 2

    work page 2017

Show all 41 references
  1. [9]

    Visibility probability structure from sfm datasets and applica- tions

    Siddharth Choudhary and PJ Narayanan. Visibility probability structure from sfm datasets and applica- tions. In Proc. ECCV, 2012. 2

    2012

  2. [10]

    Optimal randomized ransac

    Ond ˇrej Chum and Ji ˇr´ı Matas. Optimal randomized ransac. IEEE Transactions on Pattern Analysis and Machine Intelligence, 30(8):1472–1482, 2008. 6

    2008

  3. [11]

    D2-net: A trainable cnn for joint detection and description of local features

    Mihai Dusmanu, Ignacio Rocco, Tomas Pajdla, Marc Pollefeys, Josef Sivic, Akihiko Torii, and Torsten Sat- tler. D2-net: A trainable cnn for joint detection and description of local features. In Proc. CVPR, 2019. 8

    2019

  4. [12]

    Random sampling concensus: A paradigm for model fitting with applications to image analysis and automated cartography

    Martin A Fischler. Random sampling concensus: A paradigm for model fitting with applications to image analysis and automated cartography. Commun. ACM, 24(6):381–395, 1981. 1

    1981

  5. [13]

    Multi-view stereo for community photo collections

    Michael Goesele, Noah Snavely, Brian Curless, Hugues Hoppe, and Steven M Seitz. Multi-view stereo for community photo collections. In Proc. ICCV ,

  6. [14]

    From structure-from-motion point clouds to fast location recognition

    Arnold Irschara, Christopher Zach, J-M Frahm, and Horst Bischof. From structure-from-motion point clouds to fast location recognition. In Proc. CVPR,

  7. [15]

    Hamming embedding and weak geometric consis- tency for large scale image search

    Herv ´e J ´egou, Matthijs Douze, and Cordelia Schmid. Hamming embedding and weak geometric consis- tency for large scale image search. In Proc. ECCV,

  8. [16]

    On the burstiness of visual elements

    Herv ´e J ´egou, Matthijs Douze, and Cordelia Schmid. On the burstiness of visual elements. In Proc. CVPR,

  9. [17]

    Aggregating local image descriptors into compact codes

    Herv ´e J ´egou, Florent Perronnin, Matthijs Douze, Jorge S ´anchez, Patrick Perez, and Cordelia Schmid. Aggregating local image descriptors into compact codes. IEEE transactions on pattern analysis and ma- chine intelligence, 34(9):1704–1716, 2012. 3

    2012

  10. [18]

    Geometric loss functions for camera pose regression with deep learn- ing

    Alex Kendall and Roberto Cipolla. Geometric loss functions for camera pose regression with deep learn- ing. In Proc. CVPR, 2017. 1, 2

    2017

  11. [19]

    Posenet: A convolutional network for real-time 6-dof camera relocalization

    Alex Kendall, Matthew Grimes, and Roberto Cipolla. Posenet: A convolutional network for real-time 6-dof camera relocalization. In Proc. ICCV, 2015. 1, 2

    2015

  12. [20]

    A novel parametrization of the perspective- three-point problem for a direct computation of abso- lute camera position and orientation

    Laurent Kneip, Davide Scaramuzza, and Roland Sieg- wart. A novel parametrization of the perspective- three-point problem for a direct computation of abso- lute camera position and orientation. In Proc. CVPR,

  13. [21]

    Worldwide pose estimation using 3d point clouds

    Yunpeng Li, Noah Snavely, Dan Huttenlocher, and Pascal Fua. Worldwide pose estimation using 3d point clouds. In Proc. ECCV. 2012. 1, 2, 6

    2012

  14. [22]

    Location recognition using prioritized feature matching

    Yunpeng Li, Noah Snavely, and Daniel P Hutten- locher. Location recognition using prioritized feature matching. In Proc. ECCV. 2010. 2, 6

    2010

  15. [23]

    Efficient global 2d-3d matching for camera localization in a large-scale 3d map

    Liu Liu, Hongdong Li, and Yuchao Dai. Efficient global 2d-3d matching for camera localization in a large-scale 3d map. In Proc. ICCV, 2017. 1, 2, 6, 7

    2017

  16. [24]

    Distinctive image features from scale- invariant keypoints

    David G Lowe. Distinctive image features from scale- invariant keypoints. IJCV, 60(2):91–110, 2004. 2

    2004

  17. [25]

    Get out of my lab: Large-scale, real-time visual-inertial localization

    Simon Lynen, Torsten Sattler, Michael Bosse, Joel Hesch, and Roland Siegwart. Get out of my lab: Large-scale, real-time visual-inertial localization. In Robotics: Science & Systems, 2015. 2

    2015

  18. [26]

    1 year, 1000 km: The oxford robot- car dataset

    Will Maddern, Geoffrey Pascoe, Chris Linegar, and Paul Newman. 1 year, 1000 km: The oxford robot- car dataset. The International Journal of Robotics Re- search, 36(1):3–15, 2017. 7

    2017

  19. [27]

    From coarse to fine: Robust hierarchical localization at large scale

    Paul-Edouard Sarlin, Cesar Cadena, Roland Siegwart, and Marcin Dymczyk. From coarse to fine: Robust hierarchical localization at large scale. InProc. CVPR,

  20. [28]

    Leveraging deep visual descriptors for hierarchical ef- ficient localization

    Paul-Edouard Sarlin, Fr ´ed´eric Debraine, Marcin Dymczyk, Roland Siegwart, and Cesar Cadena. Leveraging deep visual descriptors for hierarchical ef- ficient localization. In Proc. CoRL, 2018. 2

    2018

  21. [29]

    Hyperpoints and fine vocabularies for large-scale location recogni- tion

    Torsten Sattler, Michal Havlena, Filip Radenovic, Konrad Schindler, and Marc Pollefeys. Hyperpoints and fine vocabularies for large-scale location recogni- tion. In Proc. ICCV, 2015. 1, 2, 4, 8

    2015

  22. [30]

    Fast image-based localization using direct 2d-to-3d match- ing

    Torsten Sattler, Bastian Leibe, and Leif Kobbelt. Fast image-based localization using direct 2d-to-3d match- ing. In Proc. ICCV, 2011. 2, 6

    2011

  23. [31]

    Im- proving image-based localization by active correspon- dence search

    Torsten Sattler, Bastian Leibe, and Leif Kobbelt. Im- proving image-based localization by active correspon- dence search. In Proc. ECCV. 2012. 1, 2, 3, 6, 7

    2012

  24. [32]

    Ef- ficient & effective prioritized matching for large-scale image-based localization

    Torsten Sattler, Bastian Leibe, and Leif Kobbelt. Ef- ficient & effective prioritized matching for large-scale image-based localization. IEEE transactions on pat- tern analysis and machine intelligence , 39(9):1744– 1756, 2017. 2

    2017

  25. [33]

    Benchmarking 6dof outdoor visual localization in changing conditions

    Torsten Sattler, Will Maddern, Carl Toft, Akihiko Torii, Lars Hammarstrand, Erik Stenborg, Daniel Sa- fari, Masatoshi Okutomi, Marc Pollefeys, Josef Sivic, Fredrik Kahl, and Tomas Pajdla. Benchmarking 6dof outdoor visual localization in changing conditions. In Proc. CVPR, 2018. 1, 6

    2018

  26. [34]

    Are large-scale 3d models really necessary for accurate visual localization? In Proc

    Torsten Sattler, Akihiko Torii, Josef Sivic, Marc Polle- feys, Hajime Taira, Masatoshi Okutomi, and Tomas Pajdla. Are large-scale 3d models really necessary for accurate visual localization? In Proc. CVPR, 2017. 1, 6

    2017

  27. [35]

    Image retrieval for image-based local- ization revisited

    Torsten Sattler, Tobias Weyand, Bastian Leibe, and Leif Kobbelt. Image retrieval for image-based local- ization revisited. In Proc. BMVC, 2012. 2, 3

    2012

  28. [36]

    Photo tourism: exploring photo collections in 3d

    Noah Snavely, Steven M Seitz, and Richard Szeliski. Photo tourism: exploring photo collections in 3d. In ACM transactions on graphics (TOG) , volume 25, pages 835–846. ACM, 2006. 1

    2006

  29. [37]

    City-scale localization for cameras with known vertical direction

    Linus Sv ¨arm, Olof Enqvist, Fredrik Kahl, and Magnus Oskarsson. City-scale localization for cameras with known vertical direction. IEEE transactions on pat- tern analysis and machine intelligence , 39(7):1455– 1461, 2017. 1, 2, 6, 7

    2017

  30. [38]

    Semantic match consistency for long-term vi- sual localization

    Carl Toft, Erik Stenborg, Lars Hammarstrand, Lucas Brynte, Marc Pollefeys, Torsten Sattler, and Fredrik Kahl. Semantic match consistency for long-term vi- sual localization. In Proc. ECCV, 2018. 1, 2, 7

    2018

  31. [39]

    On-device scal- able image-based localization via prioritized cascade search and fast one-many ransac

    Ngoc-Trung Tran, Dang-Khoa Le Tan, Anh-Dzung Doan, Thanh-Toan Do, Tuan-Anh Bui, Mengx- uan Tan, and Ngai-Man Cheung. On-device scal- able image-based localization via prioritized cascade search and fast one-many ransac. IEEE Transactions on Image Processing, 28(4):1675–1690, 2019. 2

    2019

  32. [40]

    Image-based localization using lstms for struc- tured feature correlation

    Florian Walch, Caner Hazirbas, Laura Leal-Taixe, Torsten Sattler, Sebastian Hilsenbeck, and Daniel Cre- mers. Image-based localization using lstms for struc- tured feature correlation. In Proc. ICCV, 2017. 2

    2017

  33. [41]

    Camera pose voting for large-scale image-based local- ization

    Bernhard Zeisl, Torsten Sattler, and Marc Pollefeys. Camera pose voting for large-scale image-based local- ization. In Proc. ICCV, 2015. 1, 2, 3, 6, 7

    2015

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.