pith. sign in

arxiv: 2511.15875 · v2 · submitted 2025-11-19 · 💻 cs.CV

Automatic Uncertainty-Aware Synthetic Data Bootstrapping for Historical Map Segmentation

Lukas Arzoumanidis , Julius Knechtel , Jan-Henrik Haunert , Youness Dehbi This is my paper

Pith reviewed 2026-05-17 20:05 UTC · model grok-4.3

classification 💻 cs.CV
keywords historical mapssynthetic data generationdomain adaptationsemantic segmentationaleatoric uncertaintystyle transferland cover interpretationgraph convolutional networks
0
0 comments X

The pith

Transferring historical map styles onto modern vector data creates unlimited synthetic training images that support domain-adaptive segmentation of real historical maps.

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

The paper addresses the scarcity of annotated data for analyzing specific historical map corpora by generating synthetic examples. It transfers the visual style of existing historical maps onto clean modern vector data and adds simulated noise and uncertainty to mimic real scans. This produces large, varied training sets without manual labeling. The synthetic data then trains a graph convolutional network for land-cover segmentation on actual historical maps, testing whether the emulated uncertainty closes the gap between synthetic and real domains.

Core claim

By applying cartographic style transfer from a historical map corpus to modern vector data and then emulating aleatoric uncertainty through either a deep generative model or a manual stochastic degradation process, an effectively unlimited supply of synthetic historical maps can be produced that support effective domain-adaptive semantic segmentation of real maps in the same homogeneous corpus.

What carries the argument

Cartographic style transfer from real historical maps onto modern vector data combined with emulation of visual uncertainty and noise to generate synthetic training images.

If this is right

  • Unlimited quantities of training images become available for any homogeneous historical map corpus without further manual annotation.
  • Both an automatic deep generative method and a simpler manual stochastic degradation method can produce usable synthetic data.
  • A Self-Constructing Graph Convolutional Network trained on the synthetic data can perform land-cover interpretation on real maps from the target corpus.
  • The approach directly targets the domain shift caused by aging, scanning artifacts, and cartographic conventions in historical documents.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same style-transfer-plus-uncertainty pipeline could be tested on other types of historical documents such as manuscripts or architectural drawings.
  • Combining the synthetic data with small amounts of real labeled maps might further improve accuracy beyond using either source alone.
  • The generated datasets could serve as a public resource for benchmarking domain adaptation methods in cartographic analysis.
  • If the uncertainty emulation proves transferable, the method could reduce annotation costs in large-scale digital humanities projects involving map corpora.

Load-bearing premise

Emulating visual uncertainty and noise in the synthetic images will be enough to bridge the domain gap so that models trained on them perform well on real historical map scans.

What would settle it

A model trained only on the generated synthetic maps achieves substantially lower segmentation accuracy on held-out real historical maps than a model trained on even a small set of manually labeled real maps from the same corpus.

read the original abstract

The automated analysis of historical documents, particularly maps, has drastically benefited from advances in deep learning and its success across various computer vision applications. However, most deep learning-based methods heavily rely on large amounts of annotated training data, which are typically unavailable for historical maps, especially for those belonging to specific, homogeneous cartographic domains, also known as corpora. Creating high-quality training data suitable for machine learning often takes a significant amount of time and involves extensive manual effort. While synthetic training data can alleviate the scarcity of real-world samples, it often lacks the affinity (realism) and diversity (variation) necessary for effective learning. By transferring the cartographic style of a historical map corpus onto modern vector data, we bootstrap an effectively unlimited number of synthetic historical maps suitable for tasks such as land-cover interpretation of a homogeneous historical map corpus. We propose an automatic deep generative approach and an alternative manual stochastic degradation technique to emulate the visual uncertainty and noise, also known as aleatoric uncertainty, commonly observed in historical map scans. To quantitatively evaluate the effectiveness and applicability of our approach, the bootstrapped training datasets were employed for domain-adaptive semantic segmentation on a homogeneous map corpus using a Self-Constructing Graph Convolutional Network, enabling a comprehensive assessment of the impact of our data bootstrapping methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims that transferring the cartographic style of a historical map corpus onto modern vector data, via an automatic deep generative model combined with manual stochastic degradation to emulate aleatoric uncertainty and noise, bootstraps unlimited synthetic historical maps suitable for land-cover segmentation; these are then used to train a Self-Constructing Graph Convolutional Network (SCGCN) for domain-adaptive semantic segmentation on a homogeneous historical map corpus.

Significance. If the synthetic data distribution is shown to be sufficiently close to real historical scans, the approach would address the scarcity of annotated data for specialized cartographic domains and reduce reliance on manual labeling. The combination of style transfer with explicit uncertainty emulation is a targeted contribution to synthetic data generation for document image analysis.

major comments (2)
  1. [Abstract and Evaluation] Abstract and Evaluation section: the central claim that style-transferred vectors plus emulated aleatoric uncertainty close the domain gap sufficiently for effective SCGCN segmentation rests on an untested assumption; no distribution-level metrics (FID, MMD on deep features) between synthetic and real maps are reported, and the single-corpus SCGCN evaluation does not ablate the necessity of the uncertainty emulation component.
  2. [Method] Method description: the automatic deep generative approach and the manual stochastic degradation are presented as alternatives, yet the manuscript does not quantify how well either captures the relevant scan artifacts, ink variation, or paper texture that actually affect downstream segmentation performance.
minor comments (2)
  1. [Method] Clarify the precise architecture and training details of the deep generative model used for style transfer, including any hyperparameters that control the degree of historical-map realism.
  2. [Abstract] The abstract states that quantitative evaluation was performed, yet no numerical results (accuracy, IoU, or ablation tables) appear in the provided summary; ensure these are prominently reported with statistical significance tests.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review. We address each major comment below and indicate the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract and Evaluation] Abstract and Evaluation section: the central claim that style-transferred vectors plus emulated aleatoric uncertainty close the domain gap sufficiently for effective SCGCN segmentation rests on an untested assumption; no distribution-level metrics (FID, MMD on deep features) between synthetic and real maps are reported, and the single-corpus SCGCN evaluation does not ablate the necessity of the uncertainty emulation component.

    Authors: We agree that direct distribution-level metrics would provide additional support for the claim of domain gap closure. Our evaluation centered on downstream segmentation performance using the SCGCN on the target historical corpus, which offers a task-specific measure of utility. To address this, we will add FID and MMD computations on deep features in the revised manuscript. Regarding ablation of the uncertainty emulation, the manuscript already contrasts the two proposed bootstrapping methods (both incorporating emulation) against non-synthetic baselines; however, we will include an explicit ablation removing the stochastic degradation step to isolate its contribution. revision: partial

  2. Referee: [Method] Method description: the automatic deep generative approach and the manual stochastic degradation are presented as alternatives, yet the manuscript does not quantify how well either captures the relevant scan artifacts, ink variation, or paper texture that actually affect downstream segmentation performance.

    Authors: The automatic generative model is trained directly on the historical corpus to learn and replicate its style distribution, including artifacts, while the manual stochastic method applies targeted degradations calibrated to observed historical map characteristics. We acknowledge that the manuscript does not include explicit quantitative metrics for individual artifact types. In revision we will add qualitative side-by-side comparisons and simple proxy measures (e.g., local texture statistics) to better illustrate the emulation quality. revision: partial

Circularity Check

0 steps flagged

No significant circularity; method is self-contained empirical proposal

full rationale

The paper describes an applied pipeline: style transfer from historical maps onto modern vectors, plus manual or generative emulation of aleatoric noise, followed by training a Self-Constructing Graph Convolutional Network for domain-adaptive segmentation. No equations or steps reduce by construction to fitted inputs or self-citations; the central claim is supported by external models and reported empirical results on a held-out corpus rather than by re-labeling of the training procedure itself. The derivation chain therefore remains independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract does not specify any free parameters, axioms, or invented entities; details would be in the full manuscript.

pith-pipeline@v0.9.0 · 5541 in / 992 out tokens · 129354 ms · 2026-05-17T20:05:40.903349+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    https://doi.org/10.1007/978-3-319-66908-3

    Chiang, Y.-Y., Duan, W., Leyk, S., Uhl, J.H., Knoblock, C.A.: Using Historical Maps in Scientific Studies: Applications, Challenges, and Best Practices, (2020). https://doi.org/10.1007/978-3-319-66908-3

    work page doi:10.1007/978-3-319-66908-3 2020

  2. [2]

    In: Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for 20 Geographic Knowledge Discovery

    Li, Z., Guan, R., Yu, Q., Chiang, Y.-Y., Knoblock, C.A.: Synthetic map gen- eration to provide unlimited training data for historical map text detection. In: Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for 20 Geographic Knowledge Discovery. GEOAI ’21, pp. 17–26. Association for Com- puting Machinery, New York, NY, USA (2021). https:...

    work page doi:10.1145/3486635 2021

  3. [3]

    GIScience & Remote Sensing59(1), 200–214 (2022) https: //doi.org/10.1080/15481603.2021.2023840

    Wu, S., Heitzler, M., Hurni, L.: Leveraging uncertainty estimation and spa- tial pyramid pooling for extracting hydrological features from scanned historical topographic maps. GIScience & Remote Sensing59(1), 200–214 (2022) https: //doi.org/10.1080/15481603.2021.2023840

    work page doi:10.1080/15481603.2021.2023840 2022

  4. [4]

    In: Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems

    Wu, S., Chen, Y., Schindler, K., Hurni, L.: Cross-attention spatio-temporal con- text transformer for semantic segmentation of historical maps. In: Proceedings of the 31st ACM International Conference on Advances in Geographic Information Systems. SIGSPATIAL ’23. Association for Computing Machinery, New York, NY, USA (2023). https://doi.org/10.1145/358913...

    work page doi:10.1145/3589132.3625572 2023

  5. [5]

    Car- tography and Geographic Information Science (2025) https://doi.org/10.1080/ 15230406.2025.2468304

    Arzoumanidis, L., Knechtel, J., Haunert, J.-H., Dehbi, Y.: Semantic segmentation of historical maps using self-constructing graph convolutional networks. Car- tography and Geographic Information Science (2025) https://doi.org/10.1080/ 15230406.2025.2468304

    work page arXiv 2025

  6. [6]

    Comput- ers, Environment and Urban Systems94, 101794 (2022) https://doi.org/10.1016/ j.compenvurbsys.2022.101794

    Uhl, J.H., Leyk, S., Chiang, Y.-Y., Knoblock, C.A.: Towards the automated large-scale reconstruction of past road networks from historical maps. Comput- ers, Environment and Urban Systems94, 101794 (2022) https://doi.org/10.1016/ j.compenvurbsys.2022.101794

    work page arXiv 2022

  7. [7]

    ISPRS International Journal of Geo-Information13(12) (2024) https://doi.org/10.3390/ijgi13120464

    Sertel, E., Hucko, C.M., Kabadayı, M.E.: Automatic road extraction from his- torical maps using transformer-based segformers. ISPRS International Journal of Geo-Information13(12) (2024) https://doi.org/10.3390/ijgi13120464

    work page doi:10.3390/ijgi13120464 2024

  8. [8]

    Computers, Environment and Urban Systems108, 102060 (2024) https://doi

    Jiao, C., Heitzler, M., Hurni, L.: A novel framework for road vectorization and classification from historical maps based on deep learning and symbol painting. Computers, Environment and Urban Systems108, 102060 (2024) https://doi. org/10.1016/j.compenvurbsys.2023.102060

    work page doi:10.1016/j.compenvurbsys.2023.102060 2024

  9. [9]

    Remote Sensing13(18) (2021) https://doi.org/10.3390/ rs13183672

    Uhl, J.H., Leyk, S., Li, Z., Duan, W., Shbita, B., Chiang, Y.-Y., Knoblock, C.A.: Combining remote-sensing-derived data and historical maps for long-term back- casting of urban extents. Remote Sensing13(18) (2021) https://doi.org/10.3390/ rs13183672

    work page 2021

  10. [10]

    In: Yin, X.-C., Karatzas, D., Lopresti, D

    L´ opez-Rauhut, M., Zhou, H., Aubry, M., Landrieu, L.: Segmenting france across four centuries. In: Yin, X.-C., Karatzas, D., Lopresti, D. (eds.) Document Analysis and Recognition – ICDAR 2025, pp. 3–22. Springer, Cham (2026). https://doi. org/10.1007/978-3-032-04617-8 1

    work page doi:10.1007/978-3-032-04617-8 2025

  11. [11]

    Ambio52(11), 1777–1792 (2023) https://doi.org/10.1007/s13280-023-01838-z 21

    M¨ ayr¨ a, J., Kivinen, S., Keski-Saari, S., Poikolainen, L., Kumpula, T.: Utilizing historical maps in identification of long-term land use and land cover changes. Ambio52(11), 1777–1792 (2023) https://doi.org/10.1007/s13280-023-01838-z 21

    work page doi:10.1007/s13280-023-01838-z 2023

  12. [12]

    Abstracts of the ICA7, 7 (2024) https://doi.org/10.5194/ ica-abs-7-7-2024

    Arzoumanidis, L., Fethers, J.O., Mudiyanselage, S.H., Dehbi, Y.: Deep generation of synthetic training data for the automated extraction of semantic knowledge from historical maps. Abstracts of the ICA7, 7 (2024) https://doi.org/10.5194/ ica-abs-7-7-2024

    work page 2024

  13. [13]

    ISPRS Journal of Photogrammetry and Remote Sensing197, 199–211 (2023) https://doi.org/10.1016/j.isprsjprs.2023.01.021

    Wu, S., Schindler, K., Heitzler, M., Hurni, L.: Domain adaptation in segmenting historical maps: A weakly supervised approach through spatial co-occurrence. ISPRS Journal of Photogrammetry and Remote Sensing197, 199–211 (2023) https://doi.org/10.1016/j.isprsjprs.2023.01.021

    work page doi:10.1016/j.isprsjprs.2023.01.021 2023

  14. [14]

    https://arxiv.org/abs/2510.27547

    Xia, X., Balestriero, R., Zhang, T., Zhou, Y., Ding, A., Saini, D., Hurni, L.: Map- SAM2: Adapting SAM2 for Automatic Segmentation of Historical Map Images and Time Series (2025). https://arxiv.org/abs/2510.27547

    work page arXiv 2025

  15. [15]

    https://arxiv.org/abs/ 2506.21826

    Sterzinger, R., Peer, M., Sablatnig, R.: Few-Shot Segmentation of Historical Maps via Linear Probing of Vision Foundation Models (2025). https://arxiv.org/abs/ 2506.21826

    work page arXiv 2025

  16. [16]

    https://doi.org/10.48550/arXiv.2504.11050

    Yuan, Y., Sester, M.: Leveraging LLMs and attention-mechanism for automatic annotation of historical maps (2025). https://doi.org/10.48550/arXiv.2504.11050

    work page doi:10.48550/arxiv.2504.11050 2025

  17. [17]

    In: Proceedings of the ACM SIGSPA- TIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’25), Minneapolis, MN (2025)

    Duan, W., Chiang, Y., Chen, T., Gerlek, M.P., Jang, L., Kirsanova, S., Knoblock, C.A., Lin, F., Lin, Y., Li, Z., Minton, S.N.: Digmapper: A modular system for automated geologic map digitization. In: Proceedings of the ACM SIGSPA- TIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’25), Minneapolis, MN (2025)

    work page 2025

  18. [18]

    IEEE Access12, 15642–15650 (2024) https: //doi.org/10.1109/ACCESS.2024.3356122

    Bird, J.J., Lotfi, A.: Cifake: Image classification and explainable identification of ai-generated synthetic images. IEEE Access12, 15642–15650 (2024) https: //doi.org/10.1109/ACCESS.2024.3356122

    work page doi:10.1109/access.2024.3356122 2024

  19. [19]

    Schmon, and Chris G

    Kloukiniotis, A., Papandreou, A., Anagnostopoulos, C., Lalos, A., Kapsalas, P., Nguyen, D.-V., Moustakas, K.: Carlascenes: A synthetic dataset for odometry in autonomous driving. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pp. 4520–4528 (2022). https: //doi.org/10.1109/CVPRW56347.2022.00498

    work page doi:10.1109/cvprw56347.2022.00498 2022

  20. [20]

    Kirsanova, S., Chiang, Y.-Y., Duan, W.: Detecting Legend Items on Historical Maps Using GPT-4o with In-Context Learning (2025)

    work page 2025

  21. [21]

    https://arxiv.org/abs/2410.02250

    M¨ uhlematter, D.J., Schweizer, S., Jiao, C., Xia, X., Heitzler, M., Hurni, L.: Probabilistic road classification in historical maps using synthetic data and deep learning (2024). https://arxiv.org/abs/2410.02250

    work page arXiv 2024

  22. [22]

    arXiv (2025)

    Affolter, C., Wu, S., Chen, Y., Hurni, L.: Generative AI in Map-Making: A Technical Exploration and Its Implications for Cartographers. arXiv (2025). https://doi.org/10.48550/ARXIV.2508.18959 22

    work page doi:10.48550/arxiv.2508.18959 2025

  23. [23]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp

    Zhang, L., Rao, A., Agrawala, M.: Adding conditional control to text-to-image diffusion models. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 3836–3847 (2023)

    work page 2023

  24. [24]

    ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information SciencesX-4/W5-2024, 33–39 (2024) https://doi.org/10

    Arzoumanidis, L., Hecht, J., Dehbi, Y.: Towards a deep automatic generation of figure-ground maps. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information SciencesX-4/W5-2024, 33–39 (2024) https://doi.org/10. 5194/isprs-annals-X-4-W5-2024-33-2024

    work page 2024

  25. [25]

    International Journal of Cartography, 1–15 (2025) https://doi.org/10.1080/23729333.2025.2545586

    Yuan, Y., Thiemann, F., Dahms, T., Sester, M.: Semantic segmentation of time- series of historical maps by learning from only one map. International Journal of Cartography, 1–15 (2025) https://doi.org/10.1080/23729333.2025.2545586

    work page doi:10.1080/23729333.2025.2545586 2025

  26. [26]

    Konstantin Klemmer, Esther Rolf, Caleb Robinson, Lester Mackey, and Marc Rußwurm

    Xia, X., Zhang, D., Song, W., Huang, W., and, L.H.: MapSAM: adapting segment anything model for automated feature detection in historical maps. GIScience & Remote Sensing62(1) (2025) https://doi.org/10.1080/15481603.2025.2494883

    work page doi:10.1080/15481603.2025.2494883 2025

  27. [27]

    https://arxiv.org/abs/2508.05501

    Yuan, Y., Thiemann, F., Dahms, T., Sester, M.: SMOL-MapSeg: Show Me One Label (2025). https://arxiv.org/abs/2508.05501

    work page arXiv 2025

  28. [28]

    International Journal of Geo- graphical Information Science, 1–34 (2025) https://doi.org/10.1080/13658816

    Wang, C., Kang, Y., Gong, Z., Zhao, P., Feng, Y., Zhang, W., and, G.L.: Car- toagent: a multimodal large language model-powered multi-agent cartographic framework for map style transfer and evaluation. International Journal of Geo- graphical Information Science, 1–34 (2025) https://doi.org/10.1080/13658816. 2025.2507844

    work page doi:10.1080/13658816 2025

  29. [29]

    https://www.openstreetmap.org

    OpenStreetMap contributors: OpenStreetMap: Collaborative Project to Create a Free Editable Map of the World. https://www.openstreetmap.org. Accessed: October 25, 2025 (2025)

    work page 2025

  30. [30]

    https://www.maptiler.com/

    MapTiler: MapTiler – Maps for developers. https://www.maptiler.com/. Accessed: 2025-11-14 (2025)

    work page 2025

  31. [31]

    https://maputnik.github.io/

    Maputnik: Maputnik – Open-source visual editor for the MapLibre Style Specifi- cation. https://maputnik.github.io/. Accessed: 2025-11-14 (2025)

    work page 2025

  32. [32]

    Wu, S., Heitzler, M., Hurni, L.: A closer look at segmentation uncertainty of scanned historical maps. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information SciencesXLIII-B4-2022, 189–194 (2022) https://doi.org/10.5194/isprs-archives-XLIII-B4-2022-189-2022

    work page doi:10.5194/isprs-archives-xliii-b4-2022-189-2022 2022

  33. [33]

    In: 2017 IEEE International Con- ference on Computer Vision (ICCV), pp

    Zhu, J.-Y., Park, T., Isola, P., Efros, A.A.: Unpaired image-to-image translation using cycle-consistent adversarial networks. In: 2017 IEEE International Con- ference on Computer Vision (ICCV), pp. 2242–2251 (2017). https://doi.org/10. 1109/ICCV.2017.244

    work page 2017

  34. [34]

    In: ICLR (2024)

    Kim, B., Kwon, G., Kim, K., Ye, J.C.: Unpaired image-to-image translation via 23 neural schr¨ odinger bridge. In: ICLR (2024)

    work page 2024

  35. [35]

    In: Proceedings of the 31st International Conference on Neural Information Processing Systems

    Heusel, M., Ramsauer, H., Unterthiner, T., Nessler, B., Hochreiter, S.: Gans trained by a two time-scale update rule converge to a local nash equilibrium. In: Proceedings of the 31st International Conference on Neural Information Processing Systems. NIPS’17, pp. 6629–6640 (2017)

    work page 2017

  36. [36]

    Score-cam: Score-weighted visual explanations for convolutional neural net- works

    Liu, Q., Kampffmeyer, M., Jenssen, R., Salberg, A.-B.: Multi-view self- constructing graph convolutional networks with adaptive class weighting loss for semantic segmentation. In: Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2020), pp. 199–205 (2020). https://doi.org/10.1109/CVPRW50498.2020.00030

    work page doi:10.1109/cvprw50498.2020.00030 2020

  37. [37]

    International Journal of Remote Sensing42(16), 6184–6208 (2021) https://doi.org/10.1080/01431161.2021.1936267

    Liu, Q., Kampffmeyer, M., Jenssen, R., Salberg, A.-B.: Self-constructing graph neural networks to model long-range pixel dependencies for semantic segmenta- tion of remote sensing images. International Journal of Remote Sensing42(16), 6184–6208 (2021) https://doi.org/10.1080/01431161.2021.1936267

    work page doi:10.1080/01431161.2021.1936267 2021

  38. [38]

    Automation in Construction 166, 105649 (2024) https://doi.org/10.1016/j.autcon.2024.105649

    Knechtel, J., Rottmann, P., Haunert, J.-H., Dehbi, Y.: Semantic floorplan seg- mentation using self-constructing graph networks. Automation in Construction 166, 105649 (2024) https://doi.org/10.1016/j.autcon.2024.105649

    work page doi:10.1016/j.autcon.2024.105649 2024

  39. [39]

    The Fr´ echet distance revisited and extended

    Har-Peled, S., Raichel, B.: The fr´ echet distance revisited and extended. ACM Trans. Algorithms10(1) (2014) https://doi.org/10.1145/2532646

    work page doi:10.1145/2532646 2014

  40. [40]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016) 24

    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the incep- tion architecture for computer vision. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016) 24

    work page 2016