Pith. sign in

REVIEW 4 major objections 7 minor 26 references

GAN-SLAM: Real-Time GAN Aided Floor Plan Creation Through SLAM

T0 review · 4 major / 7 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A GAN trained on simulated noisy maps can clean and complete real-time LiDAR occupancy grids well enough for floor plan drafting.

desk verdict A genuinely new application of a standard GAN to live occupancy-grid cleanup, but the evaluation leans on a wall-only ground truth and a partly circular simulator FID, so the headline claim outruns the evidence. read the letter →

arxiv 2504.19653 v1 pith:KINKOVTP submitted 2025-04-28 cs.RO cs.AIcs.LG

Leon Davies , Baihua Li , Mohamad Saada , Simon S{o}lvsten , Qinggang Meng This is my paper · ORCID
classification cs.ROcs.AIcs.LG
keywords GANoccupancygridmappingSLAMLiDARodometryimage-to-imagetranslationmapcleaningfloorplancreationsim-to-realtransfer
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 tries to establish that a generative adversarial network can be inserted directly into a LiDAR SLAM pipeline to convert the noisy, incomplete occupancy grid produced by pose estimation and scan matching into a clean, complete 2D map in real time. The proposed system, GAN-SLAM, takes a single 3D LiDAR stream, estimates 6-DOF odometry without an IMU, builds an occupancy grid, and passes it through a trained image-to-image generator that removes accidental observations, straightens linear and angular offsets, and completes rooms from partial scans. The authors argue that this makes occupancy grid maps usable for human-facing downstream tasks such as floor plan drafting, and report that the GAN improves map quality over all pure-LiDAR SLAM baselines they test, including on an unseen real building. The wider point is that generative map refinement, despite its risk of hallucination, can be a practical post-processing layer for mapping-only applications.

What carries the argument

The load-bearing object is the trained generator G, a ResNet-based encoder-decoder that performs occupancy-grid-to-occupancy-grid translation; it is trained with an adversarial PatchGAN loss plus a query-selected patchwise contrastive loss and an identity loss, so it learns which map features are errors to remove and which partial observations to complete. Working alongside it is a preprocessing stage that deletes low-confidence occupied cells and isolated floating pixels, safe only because the generator is expected to rebuild them, and a 6-DOF odometry estimate computed with generalized ICP on filtered 3D pointclouds, projected into the 2D frame so that the pose estimate and the 2D scan are aligned.

What would settle it

Take a LiDAR scan of a building with a circular or strongly curved room, run GAN-SLAM, and compare the corrected occupancy grid to a ground-truth floor plan; if the output squares off the curve or removes a real corridor, as the paper already reports happening for the Intel Lab corridor, the simulation-to-real transfer claim is falsified. A second probe: feed the generator a map with a deliberately deleted wall segment and check whether the completed wall matches the true geometry or a plausible square-room prior.

Watch

Extended reading notes

Core claim

The central claim is that occupancy grid cleaning and completion can be treated as an image-to-image translation task: a generator G is trained adversarially, with a PatchGAN discriminator and a patchwise contrastive loss, to map erroneous occupancy grids to clean, complete grids, and this trained generator is embedded in SLAM between map estimation and publication. Because the generator can reconstruct structure, the pipeline deletes low-confidence occupied cells before inference, data that would be lost in a traditional SLAM, and the GAN fills the gaps. The authors further adapt a 6-DOF LiDAR odometry method used in 3D SLAM to the 2D occupancy grid setting by computing generalized-ICP transforms between consecutive filtered pointclouds and aligning the odometry with the 2D projection. The resulting system runs in real time on a single 3D LiDAR and, in their evaluation on a real complex building, improves intersection-over-union against a manually drawn ground truth for every SLAM algorithm they paired with the GAN.

Load-bearing premise

The whole result rests on the assumption that a GAN trained exclusively on simulated maps of square indoor layouts, with errors injected by an exploration agent, will clean and complete real-world 3D-LiDAR SLAM outputs well enough that removing low-confidence cells and filling partial rooms improves map fidelity rather than inventing structures.

Editorial extensions

If this is right

  • If the central claim holds, any pure-LiDAR 2D SLAM output can be passed through the generator to improve map fidelity and completeness without changing the underlying SLAM algorithm.
  • Room completion from partial observation becomes a live SLAM feature: a robot can publish a finished-looking floor plan before it has physically visited every room, improving mapping efficiency in large buildings.
  • The preprocessing step of deleting low-confidence cells can be applied in a mapping-only context, but not for navigation, since occupancy evidence is deliberately destroyed.
  • Better pose accuracy in the 3D frame translates into cleaner 2D maps, creating a direct bridge from 3D LiDAR odometry advances to 2D occupancy grid quality.

Reading between the lines

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

  • The dependence on square, straight-walled training layouts implies a sharp boundary: curved or unusually shaped buildings will be actively harmed, and the paper's own report of hallucinating square layouts in rare designs suggests the generator learns a strong prior over indoor floor-plan geometry rather than a general error-correction rule.
  • A controlled ablation that removes the low-confidence-cell preprocessing would reveal how much of the improvement comes from the GAN's inpainting versus simply discarding noisy data, since the paper's design bundles these two operations.
  • If the approach generalizes, it points to a cheap route to floor-plan generation: pair a commodity 360-degree LiDAR with a pretrained map-cleaning generator, avoiding expensive manual surveying.
  • The corridor-removal failure in the Intel Lab sample suggests a testable failure mode at low resolutions: downsampling large environments to 256 by 256 can make genuine passages look like noise, so resolution-aware training or tiling would be a natural extension.
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

4 major / 7 minor

Summary. The paper proposes GAN-SLAM, a real-time SLAM pipeline that applies an image-to-image GAN to occupancy grid maps produced from 3D LiDAR. A generalized-ICP odometry module adapted from DLO supplies pose estimates, an RTAB-Map-style pose graph supports loop closure, and a ResNet generator with patchwise contrastive and adversarial losses is trained on simulated HouseExpo maps generated by a DRL exploration agent that mimics Gmapping errors. The authors evaluate on one real building (Haslegrave) using four SLAM baselines with and without the GAN, on qualitative Radish samples, and on simulated validation data using FID. The central claim is that GAN-SLAM substantially improves map fidelity and completeness for floor-plan-like mapping tasks.

Significance. If the claims were fully established, the contribution would be practically useful: a single-LiDAR pipeline that produces cleaner occupancy grids in real time, plus a training procedure based on synthetic data with a real-world transfer test. The consistent IoU improvement across all four baselines on the Haslegrave dataset and the real-time computational comparison are concrete strengths, and the paper is candid about failure cases and the intended mapping-only scope. The main weakness is that the real-world metric does not separate denoising from deleting non-wall content or hallucinating square layouts, and the simulator FID evaluation is substantially circular. With stronger real-world evaluation the paper would be a worthwhile empirical contribution.

major comments (4)
  1. [IV-B, Table III] The only quantitative real-world evidence is wall-only IoU on a single building, and the authors acknowledge that the ground truth 'only walls are marked as occupied space whereas in the SLAM every object is'. Under this metric, a model that deletes furniture and other non-wall occupied cells can improve IoU even if it removes legitimate map content, and hallucinated structure in unobserved rooms is not penalized if the guessed walls coincide with the ground truth. The claimed 'substantial improvements in map fidelity and quality' therefore require metrics that separate true positives from invented structure, such as per-class precision/recall or a layout-level metric, on more than one real environment, with variance estimates.
  2. [III-D, IV-C] The FID evaluation on PseudoSLAM validation data is partly circular: the GAN is trained on erroneous maps generated by the same simulator on which it is tested, so the result mostly demonstrates style matching to the simulator's clean maps rather than denoising of real SLAM outputs. The authors appropriately warn that the samples were built with the same agent, but the abstract and Section I nevertheless use these results as evidence of improved map quality. Please either remove this evidence from the support for the headline claim or evaluate on out-of-distribution simulated SLAM outputs.
  3. [III-F, Eq. (7)] The preprocessing deletes low-confidence occupied cells and isolated floating-point pixels before GAN inference, and the text states this is 'only possible due to the implemented GAN model's ability to reconstruct the broken data.' This coupling makes the net effect a combination of a heuristic filter and a learned completion model. An ablation on the real Haslegrave data, namely filter-only, GAN-only (without the filter), and filter+GAN, is needed to establish that the GAN rather than the filtering heuristics produces the reported IoU gains, and to quantify how much true wall structure is removed by the filter.
  4. [IV-A] The Radish results are purely qualitative; the authors state that there is no ground-truth 'completed' version and 'invite readers to make their own conclusions.' Given the Intel Lab failure in which a real corridor was removed, the generalisability claim for large complex environments needs a quantitative outcome, such as the fraction of environments with major structure removed or hallucinated, or a comparison against a manually annotated target for at least a subset of the Radish maps.
minor comments (7)
  1. [III, Sec. III-C] The OGM is defined as an HxW matrix, but the GAN is described as operating on HxWx3 images; specify how the single-channel occupancy grid is expanded to three channels and how the output is mapped back.
  2. [III-C, Fig. 2] The generator is named Genc/Gdec in the text but Genc/Denc in the sentence describing Fig. 2; align the notation.
  3. [Eq. (7)] The threshold is written as '<= 2' in the second case after the text says 'two or less'; specify whether a cell with exactly two same-valued neighbours is retained or removed.
  4. [Table I] The table caption promises PseudoSLAM parameters, but the table content is missing from the manuscript; include the values for FOV, range, episode counts, and error-injection settings.
  5. [IV-D] 'Structural Similarly Index Multiplier' should be 'Structural Similarity Index Measure (SSIM)'.
  6. [Throughout] Typos such as 'Extended Karlman Filtering' (Section II-A), 'apposed' (Section IV-B), and the stray backtick in the abstract ('GAN-SLAM‘') should be corrected.
  7. [References] Reference [25] is incomplete; give the full citation with year and publisher or venue, and include a dataset release link or trained-model weights to make the results reproducible.

Circularity Check

2 steps flagged · score 4.0 of 10

Simulated FID evaluation is in-distribution and the real-world IoU metric is aligned with the GAN's wall-only training target, so part of the reported improvement is by construction; the external Haslegrave benchmark prevents full circularity.

  1. fitted input called prediction [Section IV-C (PseudoSLAM Simulated Data) and Section III-D (Data)]
    "We also evaluate our SLAMs performance on simulated data. This is the evaluation data partition built by our DRL agent. To evaluate the performance of our SLAM, we compute the FID between predictions made by the trained generator on the 1000 samples of validation data, kept out of training. We compare these with a dataset of pixel perfect occupancy grid maps, also built by the simulator."

    Domain X and Domain Y are both produced by the same PseudoSLAM simulator: the paper states 'We created an erroneous copy of every sample within the HouseExpo 2D dataset ... and use this as our domain X ... Domain Y is the baseline HouseExpo 2D Dataset.' The GAN is trained with an adversarial loss whose objective is to make G(x) indistinguishable from the simulator's clean maps. The FID evaluation then compares G(validation) to clean maps 'also built by the simulator' from that same distribution. The reported FID improvement over the simulator's baseline is therefore a measure of in-distribution fit to the GAN's own training target, not an external test of map fidelity.

  2. self definitional [Section IV-B (Real-time SLAM, Table III) and Section III-D (Data)]
    "In the ground truth only walls are marked as occupied space whereas in the SLAM every object is. We decide to show these results as despite this difference our GAN demonstrates significant improvement over baseline for all methods."

    The GAN's training target, Domain Y, is the HouseExpo2D baseline dataset, i.e., clean wall-only floor plans. The real-world ground truth is also defined as wall-only. The IoU metric therefore rewards exactly the output the GAN was trained to produce: deleting non-wall objects (furniture, people, etc.) and completing unobserved rooms as square layouts. Thus part of the reported IoU improvement over the baseline SLAMs is by construction, because the evaluation measure coincides with the GAN's learned mapping. Because the input to the GAN is real LiDAR data not seen in training, the result is not fully forced; however, the metric cannot distinguish genuine denoising from task-specific hallucination, so the headline claim of improved 'map fidelity' is not independently established.

full rationale

No self-citations are load-bearing: references [16], [19]-[24] are standard external methods, and none are by the present authors. The real-world Haslegrave benchmark (Table III) and Radish qualitative results (Table II) are external to the training simulator, which is genuine independent evidence. However, the simulated FID evaluation (Table VII) is in-distribution: the GAN was trained on the same PseudoSLAM simulator's erroneous/clean pairs, so the FID improvement largely reflects successful fitting to the training distribution. Additionally, the wall-only ground truth used for the real-world IoU is aligned with the GAN's training target, making part of the improvement a metric-alignment effect rather than demonstrated general fidelity. These two issues make the central claim partially circular, but because real-world data and external SLAM baselines are used, the result is not fully forced by definition.

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

The central claim rests on the simulation-to-real transfer of the GAN, several preprocessing heuristics, and the choice of evaluation metrics. None of these are derived; they are domain assumptions. No new physical entities are introduced.

free parameters (6)
  • Voxel grid filter resolution = 0.25 m
    Downsampling of the 3D point cloud for GICP, chosen for computational efficiency; affects pose estimate and resulting map.
  • Box filter radius = 1 m
    Removes points belonging to the robot or scanning apparatus; chosen ad hoc.
  • Floating-point removal adjacency threshold = 2 (pixels)
    Eq. 7: cells with two or fewer same-valued neighbors are converted to unexplored; a heuristic that changes map content before GAN input.
  • Preprocessing intensity rounding thresholds = not specified
    Low intensity cells are removed and high intensity cells are rounded to 0 or 1; thresholds are not reported, affecting GAN input.
  • GAN training hyperparameters = lr=2e-4, beta1=0.5, beta2=0.999, 100 epochs, batch=1
    Chosen for stability; standard but affect the final generator behavior.
  • Simulator FOV and range = FOV 180, range 8 m
    Chosen to maximize map completion in DRL exploration; controls the distribution of errors in training data.
assumptions (6)
  • standard math GICP scan matching and GAN/contrastive loss definitions (Eqs. 1-6) are taken as established results.
    The paper relies on standard optimization and deep learning results without reproving them.
  • domain assumption HouseExpo 2D layouts are representative of real indoor buildings for training.
    Section III-D uses HouseExpo as the only source of clean maps.
  • domain assumption The DRL simulator with injected errors adequately mimics real SLAM failures.
    Section III-D/E; the entire training set is built on this fidelity assumption.
  • domain assumption Low-confidence cells can be deleted safely before GAN processing because the GAN reconstructs them.
    Section III-F: 'This process is only possible due to the implemented GAN model's ability to reconstruct the broken data.'
  • domain assumption FID on simulator validation data measures real map quality.
    Section IV-C; used as the primary evaluation metric on simulated data.
  • domain assumption A manually drawn ground truth marking only walls as occupied is a valid reference for IoU.
    Section IV-B; authors acknowledge the mismatch but still use IoU for comparison.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GAN-SLAM: Real-Time GAN Aided Floor Plan Creation Through SLAM." pith.science (2026). https://pith.science/paper/KINKOVTP

@misc{pith2026250419653,
  author       = {Pith},
  title        = {Pith review of: GAN-SLAM: Real-Time GAN Aided Floor Plan Creation Through SLAM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KINKOVTP}},
  note         = {Machine review of arXiv:2504.19653}
}
read the original abstract

SLAM is a fundamental component of modern autonomous systems, providing robots and their operators with a deeper understanding of their environment. SLAM systems often encounter challenges due to the dynamic nature of robotic motion, leading to inaccuracies in mapping quality, particularly in 2D representations such as Occupancy Grid Maps. These errors can significantly degrade map quality, hindering the effectiveness of specific downstream tasks such as floor plan creation. To address this challenge, we introduce our novel 'GAN-SLAM', a new SLAM approach that leverages Generative Adversarial Networks to clean and complete occupancy grids during the SLAM process, reducing the impact of noise and inaccuracies introduced on the output map. We adapt and integrate accurate pose estimation techniques typically used for 3D SLAM into a 2D form. This enables the quality improvement 3D LiDAR-odometry has seen in recent years to be effective for 2D representations. Our results demonstrate substantial improvements in map fidelity and quality, with minimal noise and errors, affirming the effectiveness of GAN-SLAM for real-world mapping applications within large-scale complex environments. We validate our approach on real-world data operating in real-time, and on famous examples of 2D maps. The improved quality of the output map enables new downstream tasks, such as floor plan drafting, further enhancing the capabilities of autonomous systems. Our novel approach to SLAM offers a significant step forward in the field, improving the usability for SLAM in mapping-based tasks, and offers insight into the usage of GANs for OGM error correction.

Figures

Figures reproduced from arXiv: 2504.19653 by the authors.

Figure 1
Figure 1. GAN-SLAM: Data pipeline, producing a high quality defect free [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. GAN-SLAM System Architecture Diagram for real-time occupancy grid mapping. Main stages: 1) Take the input of consecutive LiDAR pointclouds [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Training procedure for 2D occupancy grid cleaning through an I2I GAN, [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Samples from the dataset used to train the GAN model. Top row: [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: A real-time prediction made by GAN-SLAM of the Haslegrave [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Ground Truth Map of the Haslegrave Area TABLE V COMPUTATIONAL TIME COMPARISON From these results it can be observed that our method has a large increase in computational cost, this however is not too damaging and still is viable for real-time inference. Gmapping and He…
Figure 7
Figure 7. Figure 7: Comparison of data produced in the modified simulator and real world data. (a) Ground truth data generated within the modified simulator. (B) [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages

  1. [1]

    A survey on lidar scanning mechanisms,

    T. Raj, F. Hanim Hashim, A. Baseri Huddin, M. F. Ibrahim, and A. Hussain, “A survey on lidar scanning mechanisms,” Electronics, vol. 9, no. 5, p. 741, 2020

    work page 2020

  2. [2]

    Bun- dle adjustment—a modern synthesis,

    B. Triggs, P. F. McLauchlan, R. I. Hartley, and A. W. Fitzgibbon, “Bun- dle adjustment—a modern synthesis,” in Vision Algorithms: Theory and Practice: International Workshop on Vision Algorithms Corfu, Greece, September 21–22, 1999 Proceedings. Springer, 2000, pp. 298–372

    work page 1999

  3. [3]

    W. Hess, D. Kohler, H. Rapp and D. Andor, ”Real-time loop clo- sure in 2D LIDAR SLAM,” 2016 IEEE International Conference on Robotics and Automation (ICRA), Stockholm, Sweden,pp. 1271-1278, doi: 10.1109/ICRA.2016.7487258, 2016

    arXiv 2016

  4. [4]

    Monte carlo localisation for mobile robots,

    F. Dellaert, D. Fox, W. Burgard, and S. Thrun, “Monte carlo localisation for mobile robots,” in Proceedings 1999 IEEE International Conference on Robotics and Automation, vol. 2, 1999, pp. 1322–1328 vol.2

    work page 1999

  5. [5]

    Improved techniques for grid mapping with rao-blackwellized particle filters,

    G. Grisetti, C. Stachniss, and W. Burgard, “Improved techniques for grid mapping with rao-blackwellized particle filters,” IEEE transactions on Robotics, vol. 23, no. 1, pp. 34–46, 2007

    2007

  6. [6]

    Nwankwo, L., Rueckert, E. (2023). Understanding Why SLAM Algo- rithms Fail in Modern Indoor Environments. In: Petri ˇc, T., Ude, A., ˇZlajpah, L. (eds) Advances in Service and Industrial Robotics. RAAD

    work page 2023

  7. [7]

    ”Convergence and con- sistency analysis for extended Kalman filter based SLAM.” IEEE Transactions on robotics 23, 2007

    Huang, Shoudong, and Gamini Dissanayake. ”Convergence and con- sistency analysis for extended Kalman filter based SLAM.” IEEE Transactions on robotics 23, 2007. 10

    work page 2007

  8. [8]

    Segal, D

    A. Segal, D. Haehnel, and S. Thrun, Generalized-icp.” in Robotics: Science and Systems (RSS), 2009

    work page 2009

Show all 26 references
  1. [9]

    Loam: Lidar odometry and mapping in realtime

    J. Zhang and S. Singh, “Loam: Lidar odometry and mapping in realtime.” in Robotics: Science and systems, vol. 2, no. 9. Berkeley, CA, 2014

    2014

  2. [10]

    ”A flexible and scalable SLAM system with full 3D motion estimation.” In 2011 IEEE international symposium on safety, security, and rescue robotics, pp

    Kohlbrecher, Stefan, Oskar V on Stryk, Johannes Meyer, and Uwe Klingauf. ”A flexible and scalable SLAM system with full 3D motion estimation.” In 2011 IEEE international symposium on safety, security, and rescue robotics, pp. 155-160. IEEE, 2011

    2011

  3. [11]

    ”SLAM Toolbox: SLAM for the dynamic world.” Journal of Open Source Software 6, no

    Macenski, Steve, and Ivona Jambrecic. ”SLAM Toolbox: SLAM for the dynamic world.” Journal of Open Source Software 6, no. 61, 2783 2021

    2021

  4. [12]

    Alberto Elfes, ”Occupancy grids: A probabilistic framework for robot perception and navigation” Carnegie Mellon University, 1989

    1989

  5. [13]

    IEEE, 1985

    Hans Moravec and Alberto Elfes, ”High resolution maps from wide angle sonar”, In Proceedings of 1985 IEEE International Conference on Robotics and Automation, volume 2, pages 116–121. IEEE, 1985

    1985

  6. [14]

    Moravec, ”Sensor fusion in certainty grids for mobile robots

    Hans P. Moravec, ”Sensor fusion in certainty grids for mobile robots. In Sensor Devices and Systems for Robotics”, pages 253–276. Springer, 1989

    1989

  7. [15]

    Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain,

    T. Shan and B. Englot, “Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain,” in 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4758–4765, 2018

    2018

  8. [16]

    Di- rect lidar odometry: Fast localisation with dense point clouds,

    K. Chen, B. T. Lopez, A.-a. Agha-mohammadi, and A. Mehta, “Di- rect lidar odometry: Fast localisation with dense point clouds,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 2000–2007, 2022

    2000

  9. [17]

    Super- glue:Learning feature matching with graph neural networks,

    P.-E. Sarlin, D. DeTone, T. Malisiewicz, and A. Rabinovich, “Super- glue:Learning feature matching with graph neural networks,” in Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4938–4947, 2020

    2020

  10. [18]

    D-lc-nets: Robust denoising and loop closing networks for lidar slam in compli- cated circumstances with noisy point clouds,

    K. Liu, A. Xiao, J. Huang, K. Cui, Y . Xing, and S. Lu, “D-lc-nets: Robust denoising and loop closing networks for lidar slam in compli- cated circumstances with noisy point clouds,” IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2022, pp. 12...

    2022

  11. [19]

    Style transfer by relaxed optimal transport and self-similarity,

    N. Kolkin, J. Salavon, and G. Shakhnarovich, “Style transfer by relaxed optimal transport and self-similarity,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10 051–10 060, 2019

    2019

  12. [20]

    Contrastive learning for unpaired image-to-image translation,

    T. Park, A. A. Efros, R. Zhang, and J.-Y . Zhu, “Contrastive learning for unpaired image-to-image translation,” in European Conference on Computer Vision, 2020

    2020

  13. [21]

    Qs-attn: Query-selected attention for contrastive learning in i2i translation,

    X. Hu, X. Zhou, Q. Huang, Z. Shi, L. Sun, and Q. Li, “Qs-attn: Query-selected attention for contrastive learning in i2i translation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 18 291–18 300, 2022

    2022

  14. [22]

    Image-to-image translation with conditional adversarial networks,

    P. Isola, J.-Y . Zhu, T. Zhou, and A. A. Efros, “Image-to-image translation with conditional adversarial networks,” in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1125–1134, 2017

    2017

  15. [23]

    Houseexpo: A large-scale 2d indoor layout dataset for learning-based algorithms on mobile robots,

    L. Tingguang, H. Danny, L. Chenming, Z. Delong, W. Chaoqun, and M. Q.-H. Meng, “Houseexpo: A large-scale 2d indoor layout dataset for learning-based algorithms on mobile robots,” arXiv preprint arXiv:1903.09845, 2019

    1903 arXiv

  16. [24]

    Labb ´e, Mathieu, and Franc ¸ois Michaud. ”RTAB-Map as an open-source lidar and visual simultaneous localisation and mapping library for large- scale and long-term online operation.” Journal of field robotics 36.2: 416-446, 2019

    2019

  17. [25]

    Andrew Howard and Nicholas Roy, The Robotics Data Set Repository (Radish, 2003

    2003

  18. [2023]

    Springer, Cham

    Mechanisms and Machine Science, vol 135. Springer, Cham

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.