Pith. sign in

REVIEW 5 major objections 5 minor 28 references

Label Correction for Road Segmentation Using Road-side Cameras

T0 review · 5 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read Training on corrected label-transfer data improves road segmentation models.

desk verdict Useful winter-road dataset and a plausible label-transfer pipeline, but the headline experiment confounds label correction with frame filtering and dataset size. read the letter →

arxiv 2502.01281 v1 pith:ZRQRX3TR submitted 2025-02-03 cs.CV

Henrik Toikka , Eerik Alamikkotervo , Risto Ojala This is my paper · ORCID
classification cs.CV
keywords roadsegmentationroadsidecameraslabeltransferimageregistrationFourier-Mellintransformwinterweatherdeeplearningsemi-automaticannotation
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 proposes a semi-automatic way to build a large, weather-diverse road segmentation training set from existing roadside camera infrastructure. For each camera, a person annotates just one frame; the annotation is then transferred to all other frames, with small camera movements corrected by Fourier-Mellin image registration. The authors claim that training on this corrected label-transfer data improves the segmentation performance of several deep learning models on both the in-domain roadside test set and an out-of-domain dashcam test set, compared with training only on the single labeled frames or on naively reused labels. If this holds, it offers a cheap route to covering the winter-weather conditions that are expensive to annotate by hand.

What carries the argument

Fourier-Mellin image registration, a dense frequency-domain alignment that estimates rotation, scaling, and translation between two images via phase correlation in log-polar coordinates. The paper uses it to warp a manually drawn road label from a reference frame onto each other frame of the same camera feed, and chains multiple registrations—selecting the chain with the highest product of registration responses—when direct registration between distant frames fails. A response threshold of 0.45 filters out frames where no reliable chain is found.

What would settle it

Manually annotate a random sample of, say, 100 frames from several cameras in the dataset, compute the IoU between the transferred label and the fresh manual annotation, and check whether IoU is systematically lower on frames where snow cover or road edge changes are visible; if those frames show large label error, the static-area assumption is violated.

Watch

Extended reading notes

Core claim

The central claim is that a single manual road annotation per roadside camera, transferred across that camera's feed with frequency-domain image registration, yields training data that improves road segmentation models. The paper demonstrates this by training DeepLabv3 with a ResNet50 backbone, a frozen DINOv2 backbone with a segmentation head, and a DINOv2 linear probe on three training sets: only the manually labeled frames, all frames with the label reused unchanged, and all frames where the transferred label is corrected by Fourier-Mellin registration. On both the roadside camera test set and a winter dashcam test set, the highest Intersection-over-Union (IoU) scores were achieved by models trained with the corrected reuse data, reaching 93.50 and 95.35 IoU respectively.

Load-bearing premise

The drivable road area in each fixed camera's view stays the same over the four-month winter collection period, so a single manual annotation remains valid for every frame of that feed; snow accumulation and plowing can change the true road boundary while the transferred label stays fixed, injecting systematic error near road edges.

Editorial extensions

If this is right

  • A large winter road dataset can be assembled from thousands of existing cameras with only 927 manual annotations, lowering the cost of weather-diverse training data.
  • Models trained this way generalize beyond the roadside perspective to an on-board dashcam view, suggesting the data can support autonomous-vehicle perception.
  • The performance gain is clearest for models trained from scratch (DeepLabv3), while frozen pretrained backbones benefit less, indicating that the value of the extra data depends on the training regime.
  • The same label-transfer pipeline could be applied to other static-camera segmentation tasks, such as lane marking or sidewalk detection, where the scene geometry is similarly stable.
  • Because frames where registration fails are filtered out, the resulting dataset under-represents the most extreme low-light and heavy-weather frames, which should be considered when training for those conditions.

Reading between the lines

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

  • The assumption of a static drivable area is most fragile under snow accumulation and road plowing; a temporal refinement step that updates the transferred label when the road boundary shifts could reduce systematic edge error.
  • The registration-response product used for path selection doubles as a confidence score; it could be used to weight samples during training so that uncertain labels contribute less.
  • Applying this method across seasons, not just winter, could yield year-round training data, and the same infrastructure could be reused for other semantic classes.
  • The dashcam generalization result hints that roadside cameras, which observe the road from a high, fixed viewpoint, may be a richer and cheaper training source than the sparse on-board data the industry currently collects.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes a semi-automatic label transfer method for road segmentation using roadside cameras. For each camera, one frame is manually annotated, and the label is transferred to other frames via Fourier-Mellin image registration, with a path-chaining scheme to handle large appearance changes. The authors collect data from 927 Finnish roadside cameras over four winter months and train three segmentation models (DeepLabv3, DINOv2+Seghead, DINOv2+Linear probe) on three training sets: a small manually labeled baseline (661 images), a large set with directly reused labels (493,411 images), and a filtered set with registration-corrected labels (164,128 images). They report IoU improvements on both an in-domain roadside test set and an out-of-domain dashcam test set, concluding that the semi-automatically labeled data boosts segmentation performance.

Significance. If the central claim holds, the method offers a highly cost-effective way to build large road-segmentation training sets from existing infrastructure, covering diverse winter weather conditions with only one manual annotation per camera. The use of 927 real camera feeds and an external dashcam benchmark is a strength: the external test set provides a meaningful out-of-distribution evaluation, and the gains on it (e.g., DeepLabv3 IoU from 87.83 to 94.72 on dashcam) are substantial and not self-referential. The paper also introduces a practical transform-chaining strategy that reduces registration failures, and it honestly discusses the limitation that snow may alter the true drivable area. However, the experimental design conflates label correction with training-set size and frame filtering, so the specific benefit of the registration-based correction is not isolated. The absence of error bars and significance tests further weakens the comparisons, especially where differences are small (e.g., DINOv2+Seghead roadside IoU 91.62 vs. 91.61).

major comments (5)
  1. [III-C and Table I] The comparison between Reuse and Corrected Reuse does not isolate the effect of label correction. Corrected Reuse contains 164,128 images that passed the registration-response threshold, whereas Reuse contains 493,411 images with no filtering. The improvement in Table I (e.g., DeepLabv3 IoU 93.50 vs. 92.69) could be due to discarding low-quality or difficult frames (low-light, extreme weather) rather than to the registration transform itself. The paper should include a control: either an equal-size random subset of Reuse, or the same 164,128 frames with labels transferred without correction, to attribute the gain to the label-correction step.
  2. [III-C and Table II] The improvement of Corrected Reuse over Baseline is confounded by a roughly 250-fold increase in training-set size (661 vs. 164,128 images). Any additional data, even with imperfect labels, could plausibly improve performance. To support the abstract's claim that 'training on the semi-automatically labeled data boosted the segmentation performance,' the paper should include a learning-curve comparison or an equal-size subset of Reuse/Corrected Reuse matched to Baseline size, so that the effect of scale is separated from the effect of label quality.
  3. [IV (Results) and Tables I–II] No error bars, multiple seeds, or significance tests are reported. All conclusions rest on single training runs. Several differences are within a few tenths of an IoU point (e.g., DINOv2+Seghead roadside 91.62 vs. 91.61; DINOv2+Linear probe dashcam 93.24 vs. 92.96), and without variance estimates these may be noise. At minimum, the authors should train each configuration with at least three seeds and report mean and standard deviation, and preferably perform a paired significance test on the test sets.
  4. [III-C (Roadside camera dataset) and IV] The in-domain roadside test set appears to be labeled by the same registration-based transfer pipeline used to generate the training data. This measures consistency with the method's own geometric alignment rather than accuracy against ground truth. A small manually annotated subset of the roadside test set should be provided to validate that the transferred labels themselves are correct, especially near road boundaries where the paper acknowledges snow accumulation can make the fixed label wrong.
  5. [III-B and III-C] The registration-quality score used for filtering is defined as the product of Fourier-Mellin response values along the transform chain (Eq. 1), and the threshold of 0.45 is stated without justification or sensitivity analysis. The paper treats the product of responses as a valid proxy for registration quality, but this is an ad-hoc heuristic. The authors should either provide evidence that this score correlates with label accuracy (e.g., by comparing a subset of filtered frames against manual alignment) or report how the results change with different thresholds.
minor comments (5)
  1. [III-B] Typo: 'lightning' should be 'lighting' in the first paragraph of Section III-B.
  2. [III-B, Eq. (1)] The notation |p| in Eq. (1) is used to denote the number of elements in the path p, but it is not defined in the text. Please define it explicitly.
  3. [II-B, Ref. [4]] Reference [4] is cited as an example of feature-based image registration, but the cited paper 'The sift algorithm for fundamental frequency estimation' concerns audio fundamental-frequency estimation, not the SIFT feature descriptor. The intended reference is likely Lowe's SIFT paper; please correct this citation.
  4. [III-D and Fig. 2] Figure 2's caption states that 'cars possibly present on the road are segmented out of the label,' but no car segmentation or removal step is described in the Methods section. Please either add the corresponding description or revise the caption to match the actual pipeline.
  5. [Throughout] The name 'Dinov2' is normally capitalized as 'DINOv2'; please make the spelling consistent with the reference [26].

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the label-transfer evaluation is self-contained and uses external test sets.

full rationale

The paper's derivation chain is entirely empirical: it proposes a Fourier-Mellin label-transfer pipeline (using OpenCV's phase correlation implementation) and then measures whether training on the transferred labels improves road segmentation. There is no fitted parameter later renamed as a prediction; the 0.45 registration-response threshold is a filtering criterion, not a quantity fitted to test outcomes. The roadside test and validation splits are fixed independently of the three training sets ('The test and validation sets remain identical across experiments'), and the dashcam test set is an external benchmark previously used in [21]. Although references [14], [20], and [21] are authored by the same research group, they are used only as related work or as provenance for the external dashcam data, not as load-bearing derivations, uniqueness theorems, or ansatz justifications. The Corrected-Reuse-versus-Reuse comparison is confounded by frame filtering and training-set size, but that is an experimental-design concern, not a circularity of definition or equation. No equation-level or definitional reduction is present.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim depends on several unstated or weakly justified numerical choices (threshold 0.45, gamma 1/1.35, batch size 24, 10% sampling) and on domain assumptions about static road geometry and the validity of the registration response as a quality measure. No new entities are introduced. The methods are standard, so the free-parameter count is moderate but the sensitivity of results to these choices is not explored.

free parameters (7)
  • registration failure threshold = 0.45 (product of Fourier-Mellin responses)
    Frames with optimal chain response product below 0.45 are filtered out; this threshold determines the 164,128-frame Corrected Reuse dataset and is set by hand with no sensitivity analysis.
  • batch transform decay factor gamma = 1/1.35
    Controls the proportion of cross-batch Fourier-Mellin transforms computed; chosen for computational load without ablation.
  • batch size for transform chaining = 24 images
    Frames are divided into batches of 24 based on timestamp; all intra-batch transforms are computed and the choice is not justified.
  • max batch distance = 8 batches
    No transforms are computed between images more than 8 batches apart; this arbitrary cutoff affects path search and dataset composition.
  • frame sampling ratio = approximately 10%
    Roughly 7000 frames per camera were collected and about 10% were randomly sampled to form the final dataset; the ratio is chosen without justification.
  • manual label count per feed = 1 image per feed
    Only one frame per camera is manually labeled; the number of manual labels needed for accuracy is not explored.
  • Fourier-Mellin preprocessing hyperparameters = Hanning window, high-pass filter, log-polar binning (unspecified)
    Preprocessing parameters are not given, so registration quality depends on unstated choices.
assumptions (5)
  • domain assumption The drivable road area in each fixed camera view is stable over the collection period, so a single manual label remains valid for all frames.
    Used throughout Section III-C and IV; the authors acknowledge in the Discussion that snow can change the drivable area, so this is a stated but unquantified assumption.
  • domain assumption Camera motion can be compensated by a Euclidean transform (translation, rotation, scale); shear, lens distortion, and perspective changes are negligible.
    Section III-A assumes non-Euclidean disturbances are negligible because camera distance is large and motion is small; no lens calibration or homography check is performed.
  • ad hoc to paper The product of Fourier-Mellin responses along a transform chain is a valid proxy for registration quality.
    Section III-B uses the response product to select chains and threshold 0.45 to discard frames; no validation against manually annotated registration error is provided.
  • domain assumption Roadside test-set ground truth labels, produced by the same transfer pipeline, are accurate enough to benchmark models.
    Since only one frame per feed is manually labeled, evaluation of the in-domain test set necessarily uses transferred labels; the paper does not state this explicitly.
  • domain assumption The single manually labeled frame per feed is accurate and representative of all frames in that feed.
    Section III-C states 'A randomly chosen frame from each feed is manually labeled'; the method assumes this label is correct for the whole feed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Label Correction for Road Segmentation Using Road-side Cameras." pith.science (2026). https://pith.science/paper/ZRQRX3TR

@misc{pith2026250201281,
  author       = {Pith},
  title        = {Pith review of: Label Correction for Road Segmentation Using Road-side Cameras},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZRQRX3TR}},
  note         = {Machine review of arXiv:2502.01281}
}
read the original abstract

Reliable road segmentation in all weather conditions is critical for intelligent transportation applications, autonomous vehicles and advanced driver's assistance systems. For robust performance, all weather conditions should be included in the training data of deep learning-based perception models. However, collecting and annotating such a dataset requires extensive resources. In this paper, existing roadside camera infrastructure is utilized for collecting road data in varying weather conditions automatically. Additionally, a novel semi-automatic annotation method for roadside cameras is proposed. For each camera, only one frame is labeled manually and then the label is transferred to other frames of that camera feed. The small camera movements between frames are compensated using frequency domain image registration. The proposed method is validated with roadside camera data collected from 927 cameras across Finland over 4 month time period during winter. Training on the semi-automatically labeled data boosted the segmentation performance of several deep learning segmentation models. Testing was carried out on two different datasets to evaluate the robustness of the resulting models. These datasets were an in-domain roadside camera dataset and out-of-domain dataset captured with a vehicle on-board camera.

Figures

Figures reproduced from arXiv: 2502.01281 by the authors.

Figure 1
Figure 1. Original manual labels are shown on the left column. Center and right [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Label correction architecture. Labels are aligned based on image [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visualisation of results on the Dashcam dataset. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of results on the roadside camera dataset. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 26 canonical work pages

  1. [1]

    Idd-aw: A benchmark for safe and robust segmentation of drive scenes in unstructured traffic and adverse weather,

    F. A. Shaik, A. Reddy, N. R. Billa, K. Chaudhary, S. Manchanda, and G. Varma, “Idd-aw: A benchmark for safe and robust segmentation of drive scenes in unstructured traffic and adverse weather,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 4614–4623

    work page 2024

  2. [2]

    Acdc: The adverse conditions dataset with correspondences for semantic driving scene understanding,

    C. Sakaridis, D. Dai, and L. Van Gool, “Acdc: The adverse conditions dataset with correspondences for semantic driving scene understanding,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 10 765–10 775

    work page 2021

  3. [3]

    Image stabilization algorithms for video-surveillance applications,

    L. Marcenaro, G. Vernazza, and C. S. Regazzoni, “Image stabilization algorithms for video-surveillance applications,” in Proceedings 2001 International Conference on Image Processing (Cat. No. 01CH37205) , vol. 1. IEEE, 2001, pp. 349–352

    work page 2001

  4. [4]

    The sift algorithm for fundamental frequency estimation,

    J. Markel, “The sift algorithm for fundamental frequency estimation,” IEEE Transactions on Audio and Electroacoustics , vol. 20, no. 5, pp. 367–377, 1972

    work page 1972

  5. [5]

    Image registration with fourier-based image correlation: A comprehensive review of developments and applications,

    X. Tong, Z. Ye, Y . Xu, S. Gao, H. Xie, Q. Du, S. Liu, X. Xu, S. Liu, K. Luan et al. , “Image registration with fourier-based image correlation: A comprehensive review of developments and applications,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , vol. 12, no. 10, pp. 4062–4081, 2019

    work page 2019

  6. [6]

    Rope3d: The roadside perception dataset for autonomous driving and monocular 3d object detection task,

    X. Ye, M. Shu, H. Li, Y . Shi, Y . Li, G. Wang, X. Tan, and E. Ding, “Rope3d: The roadside perception dataset for autonomous driving and monocular 3d object detection task,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 21 341–21 350

    work page 2022

  7. [7]

    Tumtraf event: Calibration and fusion resulting in a dataset for roadside event-based and rgb cameras,

    C. Creß, W. Zimmer, N. Purschke, B. N. Doan, S. Kirchner, V . Lakshmi- narasimhan, L. Strand, and A. C. Knoll, “Tumtraf event: Calibration and fusion resulting in a dataset for roadside event-based and rgb cameras,” IEEE Transactions on Intelligent V ehicles , 2024

    work page 2024

  8. [8]

    Traffic monitoring and vehicle tracking using roadside cameras,

    Y .-J. Wu, F.-L. Lian, and T.-H. Chang, “Traffic monitoring and vehicle tracking using roadside cameras,” in 2006 IEEE International Confer- ence on Systems, Man and Cybernetics , vol. 6. IEEE, 2006, pp. 4631– 4636

    work page 2006

Show all 28 references
  1. [9]

    High accuracy traffic monitoring using road-side line-scan cameras,

    D. Douxchamps, B. Macq, and K. Chihara, “High accuracy traffic monitoring using road-side line-scan cameras,” in 2006 IEEE Intelligent Transportation Systems Conference . IEEE, 2006, pp. 875–878

    2006

  2. [10]

    Fully automatic roadside camera calibration for traffic surveillance,

    M. Dubsk ´a, A. Herout, R. Jur ´anek, and J. Sochor, “Fully automatic roadside camera calibration for traffic surveillance,” IEEE Transactions on Intelligent Transportation Systems , vol. 16, no. 3, pp. 1162–1171, 2014

    2014

  3. [11]

    Dynamic camera calibration of roadside traffic management cameras for vehicle speed estimation,

    T. N. Schoepflin and D. J. Dailey, “Dynamic camera calibration of roadside traffic management cameras for vehicle speed estimation,” IEEE Transactions on Intelligent Transportation Systems , vol. 4, no. 2, pp. 90–98, 2003

    2003

  4. [12]

    Real-time full-stack traffic scene perception for autonomous driving with roadside cameras,

    Z. Zou, R. Zhang, S. Shen, G. Pandey, P. Chakravarty, A. Parchami, and H. X. Liu, “Real-time full-stack traffic scene perception for autonomous driving with roadside cameras,” in 2022 International Conference on Robotics and Automation (ICRA) . IEEE, 2022, pp. 890–896

    2022

  5. [13]

    Evaluating roadside perception for autonomous vehicles: Insights from field testing,

    R. Zhang, D. Meng, S. Shen, T. Wang, T. Karir, M. Maile, and H. X. Liu, “Evaluating roadside perception for autonomous vehicles: Insights from field testing,” arXiv preprint arXiv:2401.12392 , 2024

    2024 arXiv

  6. [14]

    Road surface friction estimation for winter conditions utilising general visual features,

    R. Ojala and E. Alamikkotervo, “Road surface friction estimation for winter conditions utilising general visual features,” arXiv preprint arXiv:2404.16578, 2024

    2024 arXiv

  7. [15]

    Integration of roadside camera images and weather data for monitoring winter road surface conditions,

    J. Carrillo and M. Crowley, “Integration of roadside camera images and weather data for monitoring winter road surface conditions,” arXiv preprint arXiv:2009.12165, 2020

    2009 arXiv

  8. [16]

    Automatic road segmentation of traffic images

    C.-Y . Fang, H.-P. Chou, J. M. Wang, and S.-W. Chen, “Automatic road segmentation of traffic images.” in VISAPP (2) , 2015, pp. 469–477

    2015

  9. [17]

    Winter adverse driving dataset for autonomy in inclement winter weather,

    A. M. Kurup and J. P. Bos, “Winter adverse driving dataset for autonomy in inclement winter weather,” Optical Engineering , vol. 62, no. 3, pp. 031 207–031 207, 2023

    2023

  10. [18]

    Ithaca365: Dataset and driving perception under repeated and challenging weather conditions,

    C. A. Diaz-Ruiz, Y . Xia, Y . You, J. Nino, J. Chen, J. Monica, X. Chen, K. Luo, Y . Wang, M. Emond et al. , “Ithaca365: Dataset and driving perception under repeated and challenging weather conditions,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Patter...

    2022

  11. [19]

    Semantic segmentation for road surface detection in snowy environment,

    S. Vachmanus, A. A. Ravankar, T. Emaru, and Y . Kobayashi, “Semantic segmentation for road surface detection in snowy environment,” in 2020 59th Annual Conference of the Society of Instrument and Control Engineers of Japan (SICE) . IEEE, 2020, pp. 1381–1386

    2020

  12. [20]

    Tadap: Trajectory-aided drivable area auto-labeling with pretrained self- supervised features in winter driving conditions,

    E. Alamikkotervo, R. Ojala, A. Sepp ¨anen, and K. Tammi, “Tadap: Trajectory-aided drivable area auto-labeling with pretrained self- supervised features in winter driving conditions,” IEEE Transactions on Intelligent V ehicles, 2024

    2024

  13. [21]

    Trajectory- based road autolabeling with lidar-camera fusion in winter conditions,

    E. Alamikkotervo, H. Toikka, K. Tammi, and R. Ojala, “Trajectory- based road autolabeling with lidar-camera fusion in winter conditions,” arXiv preprint arXiv:2412.02370 , 2024

    2024 arXiv

  14. [22]

    Learning off-road terrain traversability with self-supervisions only,

    J. Seo, S. Sim, and I. Shim, “Learning off-road terrain traversability with self-supervisions only,” IEEE Robotics and Automation Letters , vol. 8, no. 8, pp. 4617–4624, 2023

    2023

  15. [23]

    Self-supervised traversability prediction by learning to reconstruct safe terrain,

    R. Schmid, D. Atha, F. Sch ¨oller, S. Dey, S. Fakoorian, K. Otsu, B. Ridge, M. Bjelonic, L. Wellhausen, M. Hutter et al., “Self-supervised traversability prediction by learning to reconstruct safe terrain,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Sy...

    2022

  16. [24]

    V-strong: Visual self-supervised traversability learning for off-road navigation,

    S. Jung, J. Lee, X. Meng, B. Boots, and A. Lambert, “V-strong: Visual self-supervised traversability learning for off-road navigation,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 1766–1773

    2024

  17. [25]

    Rethinking atrous convolution for semantic image segmen- tation,

    L.-C. Chen, “Rethinking atrous convolution for semantic image segmen- tation,” arXiv preprint arXiv:1706.05587 , 2017

    2017 arXiv

  18. [26]

    Dinov2: Learning robust visual features without supervision,

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khalidov, P. Fernandez, D. Haziza, F. Massa, A. El-Nouby et al. , “Dinov2: Learning robust visual features without supervision,” arXiv preprint arXiv:2304.07193, 2023

    2023 arXiv

  19. [27]

    Stronger fewer & superior: Harnessing vision foundation models for domain generalized semantic segmentation,

    Z. Wei, L. Chen, Y . Jin, X. Ma, T. Liu, P. Ling, B. Wang, H. Chen, and J. Zheng, “Stronger fewer & superior: Harnessing vision foundation models for domain generalized semantic segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- ti...

    2024

  20. [28]

    Symmetric phase-only matched filtering of fourier-mellin transforms for image registration and recognition,

    Q.-s. Chen, M. Defrise, and F. Deconinck, “Symmetric phase-only matched filtering of fourier-mellin transforms for image registration and recognition,” IEEE Transactions on pattern analysis and machine intelligence, vol. 16, no. 12, pp. 1156–1168, 1994

    1994

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.