Pith. sign in

REVIEW 4 major objections 4 minor 109 references

Rethinking Semi-supervised Segmentation Beyond Accuracy: Reliability and Robustness

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Semi-supervised segmentation's best method, UniMatchV2, gains accuracy and robustness at the price of calibration and uncertainty quality, and a new single-score metric, RSS, is proposed to surface that trade-off.

desk verdict Useful evaluation study of UniMatchV2 with a simple composite metric; the reliability-vs-accuracy trade-off is real, but the paper overstates calibration problems and needs threshold sensitivity analysis. read the letter →

arxiv 2506.05917 v1 pith:OORQINCG submitted 2025-06-06 cs.CV cs.AIcs.LG

Steven Landgraf , Markus Hillemann , Markus Ulrich This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords semi-supervisedlearningsemanticsegmentationreliabilitycalibrationuncertaintyquantificationrobustnessexpectederrorReliableScore
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

Semi-supervised semantic segmentation is usually judged by a single number, mean Intersection over Union, which says nothing about whether a model knows when it is wrong. This paper tries to establish that the field's accuracy gains come at a hidden cost: the current state-of-the-art method, UniMatchV2, is less well calibrated and worse at flagging its own errors than a fully supervised baseline, even while it wins on accuracy and robustness. To make the trade-off visible, the authors introduce the Reliable Segmentation Score (RSS), the harmonic mean of accuracy, calibration error, and two conditional uncertainty probabilities, so a low value in any component pulls the whole score down. Across Cityscapes, Pascal VOC2012, and weather-degraded out-of-domain sets, the supervised baseline wins on RSS on Cityscapes while UniMatchV2 wins on Pascal VOC2012, showing the trade-off is real but not universal. The authors read these results as a reason to rethink evaluation protocols for safety-critical deployment, where an overconfident wrong answer is worse than a cautious one.

What carries the argument

The load-bearing object is the Reliable Segmentation Score (RSS), defined as the harmonic mean of four components, accuracy (mIoU), calibration ($1-\mathrm{ECE}$), and the two conditional uncertainty probabilities $p(\mathrm{acc}|\mathrm{cer})$ (accurate given certain) and $p(\mathrm{unc}|\mathrm{inacc})$ (uncertain given inaccurate), with equal weights as the default. Because a harmonic mean is pulled down by its smallest term, RSS does what an arithmetic average would not: it lets the weakest reliability property dominate the score, so a model that is accurate but overconfident cannot hide the deficiency. The uncertainty probabilities are computed from the Shannon entropy of the full softmax distribution, with pixels split into certain and uncertain at the median entropy of each image. The paper argues that this distribution-wide view is the necessary complement to mIoU and ECE, which read only the top softmax probability or the hard decision and therefore cannot see whether the model's output is peaked or spread, the information that reveals whether a wrong answer carries low uncertainty.

What would settle it

Recompute the reliability tables on Cityscapes at the 1/16 label split using absolute uncertainty thresholds, the median entropy of the pooled validation set or fixed cutoffs at 0.5, 1.0, and 2.0 nats, instead of each image's own median, and check whether the supervised ViT-B baseline still beats UniMatchV2 on $p(\mathrm{unc}|\mathrm{inacc})$ and on RSS; if the ordering flips, the central trade-off claim is an artifact of the threshold rule, and if it stands, the claim is threshold-independent.

Watch

Extended reading notes

Core claim

The paper's central claim, stated in its conclusion, is that "UniMatchV2 achieves superior predictive performance and robustness but is often less calibrated and produces less reliable uncertainty estimates than its supervised counterpart." The evidence for it comes from training UniMatchV2, its predecessor UniMatchV1, and a supervised ViT baseline under identical configurations and comparing them on two in-domain benchmarks, Cityscapes and Pascal VOC2012, plus fog- and rain-degraded versions of Cityscapes for out-of-domain robustness. The pattern that carries the argument is the split between accuracy and error-awareness: UniMatchV2 holds the highest mIoU at nearly every label regime, up to 0.903 on Pascal VOC2012, while its $p(\mathrm{unc}|\mathrm{inacc})$ value, the probability that a wrong prediction is flagged as uncertain, sinks as low as 0.515 on Cityscapes against the supervised baseline's 0.698. UniMatchV2 is accordingly confident about mistakes that a supervised model would hedge, and this shows up even in the qualitative examples, where it labels regions with low uncertainty where the supervised model is cautious. Fusing accuracy, calibration, and these uncertainty probabilities into the harmonic-mean RSS makes the trade-off legible as a single number, and on Cityscapes the supervised baseline tops the RSS at every label fraction.

Load-bearing premise

The load-bearing premise is that dividing each image's pixels into certain and uncertain at that image's median entropy yields meaningful reliability measurements; if a different rule, say a fixed entropy cutoff applied across the whole dataset, changed the measured $p(\mathrm{unc}|\mathrm{inacc})$ values, the reported gap between the models, and with it the paper's main claim, could shrink or reverse.

Editorial extensions

If this is right

  • If RSS becomes the evaluation standard, a semi-supervised method can no longer claim meaningful progress on a benchmark by raising mIoU alone; it must keep calibration and uncertainty quality from degrading at the same time, which the supplementary training curves show UniMatchV2 currently does not.
  • On Cityscapes the fully supervised ViT-B baseline achieves the highest RSS at every label fraction despite losing on mIoU, so accuracy-focused leaderboards and reliability-focused leaderboards would rank the same methods in opposite orders.
  • On Pascal VOC2012 UniMatchV2 leads RSS at every label count, so the accuracy-reliability trade-off is benchmark-dependent rather than a general property of semi-supervised learning.
  • Under fog and rain, UniMatchV2's RSS degrades less than the supervised baseline's, a robustness win, but its $p(\mathrm{unc}|\mathrm{inacc})$ stays near 0.6 versus roughly 0.85, meaning out-of-domain accuracy gains do not come with out-of-domain error awareness.

Reading between the lines

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

  • The per-image median threshold makes 'certain' relative to each image, so a model whose consistency training compresses its entire entropy distribution downward can look certain about errors that an absolute threshold would flag; rerunning the comparison with absolute thresholds is the cleanest test of whether the reported reliability gap is a property of the models or of the measurement rule.
  • The equal-weight RSS is a policy choice, not a discovery: weighting $p(\mathrm{unc}|\mathrm{inacc})$ more heavily than mIoU, which a safety-critical practitioner plausibly would, would push rankings even farther toward the supervised baseline.
  • A testable extension the authors do not run: add an uncertainty-aware or calibration-aware term to UniMatchV2's training objective, such as penalizing confident errors on pseudo-labeled pixels, and check whether the $p(\mathrm{unc}|\mathrm{inacc})$ gap closes without sacrificing mIoU; if it closes, the accuracy-reliability trade-off is an optimization artifact rather than an inherent cost of using
  • Adopting RSS as a community metric would change what counts as a contribution in semi-supervised segmentation: methods that are polished on accuracy alone would need to demonstrate reliability too, which would likely redirect effort toward calibration-aware pseudo-labeling and uncertainty-regularized consistency training.
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

4 major / 4 minor

Summary. The paper argues that semi-supervised semantic segmentation evaluation has focused too narrowly on mIoU and ignores reliability (calibration, uncertainty quality) and robustness. It introduces the Reliable Segmentation Score (RSS), a harmonic mean of mIoU, 1-ECE, p(acc|cer), and p(unc|inacc). The authors evaluate UniMatchV2 against UniMatchV1 and a supervised baseline on Cityscapes and Pascal VOC2012 across label fractions, plus out-of-domain tests on Foggy and Rainy Cityscapes. They report that UniMatchV2 achieves higher mIoU and robustness but is often less calibrated and produces less reliable uncertainty estimates than its supervised counterpart, concluding that evaluation protocols should adopt holistic metrics like RSS.

Significance. If validated, this is a useful contribution: it identifies a genuine blind spot in semi-supervised segmentation evaluation, proposes a simple and interpretable composite metric, and provides a first reliability-oriented comparison of a current state-of-the-art method. The out-of-domain robustness analysis is a valuable addition. However, the central empirical claim depends on a metric definition that contains a concrete error, on single-run experiments without error bars, and on a thresholding choice whose sensitivity is not examined. The paper's own Table 1 partially contradicts the claim that UniMatchV2 is 'often less calibrated'. These issues must be addressed before the conclusions can be fully supported.

major comments (4)
  1. [Section 3, Eq. (4)] Equation (4) defines RSS with omega_3 appearing in both uncertainty terms and no omega_4, although the text states there are four weights and gives the example omega_1 = 1.0, omega_2 = omega_3 = omega_4 = 1/3. This makes the proposed metric ambiguous and not reproducible as written. Please correct the equation and state clearly which weighting was used to compute the RSS values in Tables 1-3.
  2. [Section 5.1, Table 1] The conclusion that UniMatchV2 is 'often less calibrated' than its supervised counterpart is not supported by the ViT-B results in Table 1. On Cityscapes, UniMatchV2 ViT-B has ECE equal to or lower than Supervised ViT-B on four of five splits (0.019 vs 0.020, 0.025 vs 0.025, 0.015 vs 0.015, 0.011 vs 0.011); on Pascal VOC2012 it is lower on four of five splits (0.030 vs 0.038, 0.022 vs 0.026, 0.018 vs 0.018, 0.016 vs 0.017). The 'less calibrated' claim holds mainly for ViT-S on Cityscapes. Please qualify the conclusion per encoder or revise it.
  3. [Section 4.3, Tables 1-3] All quantitative results are reported from single runs without error bars, multiple seeds, or statistical significance tests. Because the central claim is a reliability/accuracy trade-off and some RSS differences are small (e.g., Table 1, Pascal VOC 1464: RSS 0.962 for UniMatchV2 ViT-B vs 0.958 for Supervised ViT-B), the reader cannot tell whether the observed rankings are stable. Please provide variance estimates from at least three seeds, or explicitly justify why single runs are sufficient.
  4. [Section 4.3, p(unc|inacc) definition] The uncertainty threshold is defined as the per-image median entropy, following Landgraf et al. [39]. This makes p(unc|inacc) a relative measure: each image is split into exactly 50% certain and 50% uncertain pixels, so the metric records whether errors fall in the higher-entropy half of each image, not whether the model is uncertain on an absolute scale. The paper does not test whether the RSS ranking (e.g., Supervised ViT-B above UniMatchV2 on Cityscapes in Table 1) is robust to alternative threshold choices, such as an absolute entropy cutoff or a dataset-level quantile. Since p(unc|inacc) strongly drives the RSS differences and the central reliability claim, please report a sensitivity analysis for the threshold.
minor comments (4)
  1. [Supplementary Material, Table 2/3 caption] There is a typo: 'Reminiscent of the finings' should read 'findings'.
  2. [Figure 2] Figure 2 is described as radar charts, but the plots appear to be line or marker charts on a radial layout; please clarify the description or correct the figure.
  3. [Section 3] The claim that the four RSS components are 'largely orthogonal' is asserted without quantitative support. If, for example, ECE and p(acc|cer) are strongly correlated, the harmonic mean may not capture four independent aspects; a correlation analysis over the evaluated models would strengthen the justification.
  4. [Section 4.3] The text says ECE is computed with the default 15-bin discretization via torchmetrics, but the number of pixels per bin and the handling of the maximum confidence bin are not described; please give the exact binning details for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: RSS is a defined metric and the central reliability claim is an empirical evaluation, not a fitted or self-referential prediction.

full rationale

The paper introduces RSS as a definitional combination of mIoU, ECE, p(acc|cer), and p(unc|inacc) (Eq. 4); it does not derive any of these components from RSS, so there is no self-definitional loop. UniMatchV2 and the supervised baseline are evaluated on their actual predictions using standard metrics (mIoU, ECE, Shannon entropy), and no parameter is fitted to the data in order to produce the reported rankings. The only self-citation that affects the protocol is the per-image median-entropy threshold suggested by Landgraf et al. [39]. That threshold is a convention inherited from the authors' prior work, not a parameter fitted to the present results, and the paper's conclusion does not reduce to the threshold by construction; the robustness of the ranking to alternative thresholds is a methodological sensitivity concern, not a circularity. The RSS weights are explicitly hand-chosen equal weights, and the paper does not claim they are derived from first principles. The central claims are therefore self-contained empirical findings rather than predictions whose outputs are identical to their inputs.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

No physical entities are introduced. The main invented construct is the RSS metric itself, which is a composite score rather than an entity with independent falsifiable evidence. The free parameters are the RSS weights and the median-entropy uncertainty threshold; both affect the reported numbers but neither is fitted to optimize the conclusion.

free parameters (2)
  • RSS component weights = Proposed equal weighting, but Eq. (4) is ambiguous (omega_3 appears twice, no omega_4)
    The weights omega_i are application-specific and chosen by hand; no principled derivation is given, and the typo in Eq. (4) makes the exact values used in the tables unclear.
  • Uncertainty threshold = Per-image median Shannon entropy
    The certain/uncertain split for p(acc|cer) and p(unc|inacc) is set by the median entropy within each image, following Landgraf et al. [39]. This choice is not fitted to the data but is a free methodological parameter that affects all uncertainty metrics.
assumptions (3)
  • domain assumption Ground-truth labels in Cityscapes, Pascal VOC, Foggy Cityscapes, and Rainy Cityscapes are correct enough to serve as references for pixel-wise reliability evaluation.
    The entire evaluation of accuracy, calibration, and uncertainty assumes these annotations are reliable references; the paper does not discuss label noise.
  • domain assumption Shannon entropy of the softmax distribution is a meaningful pixel-wise uncertainty measure and median entropy is a sensible certain/uncertain threshold.
    Section 4.3 uses entropy and the median threshold from [39]; if another uncertainty measure were used, p(acc|cer) and p(unc|inacc) could change.
  • ad hoc to paper The four components of RSS are sufficiently independent that a harmonic mean is an appropriate aggregation.
    Section 3 argues orthogonality qualitatively, but no quantitative independence analysis is provided; the choice of harmonic mean and equal weights is a design decision.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rethinking Semi-supervised Segmentation Beyond Accuracy: Reliability and Robustness." pith.science (2026). https://pith.science/paper/OORQINCG

@misc{pith2026250605917,
  author       = {Pith},
  title        = {Pith review of: Rethinking Semi-supervised Segmentation Beyond Accuracy: Reliability and Robustness},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OORQINCG}},
  note         = {Machine review of arXiv:2506.05917}
}
read the original abstract

Semantic segmentation is critical for scene understanding but demands costly pixel-wise annotations, attracting increasing attention to semi-supervised approaches to leverage abundant unlabeled data. While semi-supervised segmentation is often promoted as a path toward scalable, real-world deployment, it is astonishing that current evaluation protocols exclusively focus on segmentation accuracy, entirely overlooking reliability and robustness. These qualities, which ensure consistent performance under diverse conditions (robustness) and well-calibrated model confidences as well as meaningful uncertainties (reliability), are essential for safety-critical applications like autonomous driving, where models must handle unpredictable environments and avoid sudden failures at all costs. To address this gap, we introduce the Reliable Segmentation Score (RSS), a novel metric that combines predictive accuracy, calibration, and uncertainty quality measures via a harmonic mean. RSS penalizes deficiencies in any of its components, providing an easy and intuitive way of holistically judging segmentation models. Comprehensive evaluations of UniMatchV2 against its predecessor and a supervised baseline show that semi-supervised methods often trade reliability for accuracy. While out-of-domain evaluations demonstrate UniMatchV2's robustness, they further expose persistent reliability shortcomings. We advocate for a shift in evaluation protocols toward more holistic metrics like RSS to better align semi-supervised learning research with real-world deployment needs.

Figures

Figures reproduced from arXiv: 2506.05917 by the authors.

Figure 1
Figure 1. Qualitative comparison between the supervised and UniMatchV2 models [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Quantitative comparison of the predictive performance (mIoU), calibra [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison between the supervised and UniMatchV2 models [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Evolution of the mean Intersection over Union (mIoU) and Reliable Seg [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

109 extracted references · 59 canonical work pages

  1. [39]

    tm-Technisches Messen (2025)

    Landgraf, S., Hillemann, M., Kapler, T., Ulrich, M.: A comparative study on multi-task uncertainty quantification in semantic segmentation and monocular depth estimation. tm-Technisches Messen (2025)

    work page 2025

  2. [1]

    Information fusion76, 243–297 (2021)

    Abdar, M., Pourpanah, F., Hussain, S., Rezazadegan, D., Liu, L., Ghavamzadeh, M., Fieguth, P., Cao, X., Khosravi, A., Acharya, U.R., et al.: A review of uncer- tainty quantification in deep learning: Techniques, applications and challenges. Information fusion76, 243–297 (2021)

    2021

  3. [2]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Alonso, I., Sabater, A., Ferstl, D., Montesano, L., Murillo, A.C.: Semi-supervised semantic segmentation with pixel-level contrastive learning from a class-wise memory bank. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 8219–8228 (2021)

    2021

  4. [3]

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

    Arnab, A., Miksik, O., Torr, P.H.: On the robustness of semantic segmentation models to adversarial attacks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 888–897 (2018)

    2018

  5. [4]

    IEEE transactions on pattern analysis and machine intelligence40(4), 834–848 (2017)

    Chen, L.C., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.L.: Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE transactions on pattern analysis and machine intelligence40(4), 834–848 (2017)

    2017

  6. [5]

    In: International conference on machine learning

    Chen, T., Kornblith, S., Norouzi, M., Hinton, G.: A simple framework for con- trastive learning of visual representations. In: International conference on machine learning. pp. 1597–1607. PMLR (2020)

    2020

  7. [6]

    Advances in neural in- formation processing systems33, 22243–22255 (2020)

    Chen, T., Kornblith, S., Swersky, K., Norouzi, M., Hinton, G.E.: Big self- supervised models are strong semi-supervised learners. Advances in neural in- formation processing systems33, 22243–22255 (2020)

    2020

  8. [7]

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

    Chen, X., Yuan, Y., Zeng, G., Wang, J.: Semi-supervised semantic segmentation with cross pseudo supervision. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 2613–2622 (2021)

    2021

Show all 109 references
  1. [8]

    In: Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition

    Chen, X., He, K.: Exploring simple siamese representation learning. In: Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 15750–15758 (2021)

    2021

  2. [9]

    In: 2021 IEEE International Conference on Image Processing (ICIP)

    Chen, Y., Ouyang, X., Zhu, K., Agam, G.: Complexmix: Semi-supervised semantic segmentation via mask-based data augmentation. In: 2021 IEEE International Conference on Image Processing (ICIP). pp. 2264–2268. IEEE (2021)

    2021

  3. [10]

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

    Cordts, M., Omran, M., Ramos, S., Rehfeld, T., Enzweiler, M., Benenson, R., Franke, U., Roth, S., Schiele, B.: The cityscapes dataset for semantic urban scene understanding. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3213–3223 (2016)

    2016

  4. [11]

    In: European Conference on Computer Vision

    Croce, F., Singh, N.D., Hein, M.: Towards reliable evaluation and fast training of robust semantic segmentation models. In: European Conference on Computer Vision. pp. 180–197. Springer (2024)

    2024

  5. [12]

    De Jorge, P., Volpi, R., Torr, P.H., Rogez, G.: Reliability in semantic segmenta- tion: Are we on the right track? In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7173–7182 (2023)

    2023

  6. [13]

    Journal of Open Source Software7(70), 4101 (2022)

    Detlefsen, N.S., Borovec, J., Schock, J., Jha, A.H., Koker, T., Di Liello, L., Stancl, D., Quan, C., Grechkin, M., Falcon, W.: Torchmetrics-measuring reproducibility in pytorch. Journal of Open Source Software7(70), 4101 (2022)

    2022

  7. [14]

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

    Ding, Z., Han, X., Liu, P., Niethammer, M.: Local temperature scaling for proba- bility calibration. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 6889–6899 (2021)

    2021

  8. [15]

    International journal of computer vision111, 98–136 (2015) 14 S

    Everingham, M., Eslami, S.A., Van Gool, L., Williams, C.K., Winn, J., Zisser- man, A.: The pascal visual object classes challenge: A retrospective. International journal of computer vision111, 98–136 (2015) 14 S. Landgraf et al

    2015

  9. [16]

    Pattern Recognition130, 108777 (2022)

    Feng, Z., Zhou, Q., Gu, Q., Tan, X., Cheng, G., Lu, X., Shi, J., Ma, L.: Dmt: Dynamic mutual training for semi-supervised learning. Pattern Recognition130, 108777 (2022)

    2022

  10. [17]

    arXiv preprint arXiv:1906.01916 (2019)

    French, G., Laine, S., Aila, T., Mackiewicz, M., Finlayson, G.: Semi-supervised semantic segmentation needs strong, varied perturbations. arXiv preprint arXiv:1906.01916 (2019)

    2019 arXiv

  11. [18]

    Advances in neural information processing systems27(2014)

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial nets. Advances in neural information processing systems27(2014)

    2014

  12. [19]

    arXiv preprint arXiv:1412.6572 (2014)

    Goodfellow, I.J., Shlens, J., Szegedy, C.: Explaining and harnessing adversarial examples. arXiv preprint arXiv:1412.6572 (2014)

    2014 arXiv

  13. [20]

    In: International Conference on Machine Learning

    Guo, C., Pleiss, G., Sun, Y., Weinberger, K.Q.: On calibration of modern neu- ral networks. In: International Conference on Machine Learning. pp. 1321–1330. PMLR (2017)

    2017

  14. [21]

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

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recogni- tion. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016)

    2016

  15. [22]

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

    He, R., Yang, J., Qi, X.: Re-distributing biased pseudo labels for semi- supervised semantic segmentation: A baseline investigation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 6930–6940 (2021)

    2021

  16. [23]

    arXiv preprint arXiv:1903.12261 (2019)

    Hendrycks, D., Dietterich, T.: Benchmarking neural network robustness to com- mon corruptions and perturbations. arXiv preprint arXiv:1903.12261 (2019)

    2019 arXiv

  17. [24]

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

    Hendrycks, D., Zhao, K., Basart, S., Steinhardt, J., Song, D.: Natural adversarial examples. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 15262–15271 (2021)

    2021

  18. [25]

    arXiv preprint arXiv:1207.0580 (2012)

    Hinton, G.E., Srivastava, N., Krizhevsky, A., Sutskever, I., Salakhutdinov, R.R.: Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580 (2012)

    2012 arXiv

  19. [26]

    In: Eu- ropean Conference on Computer Vision

    Howlader, P., Das, S., Le, H., Samaras, D.: Beyond pixels: Semi-supervised se- mantic segmentation with a multi-scale patch-based multi-label classifier. In: Eu- ropean Conference on Computer Vision. pp. 342–360. Springer (2024)

    2024

  20. [27]

    In: European Conference on Computer Vision

    Hoyer, L., Tan, D.J., Naeem, M.F., Van Gool, L., Tombari, F.: Semivl: semi- supervised semantic segmentation with vision-language guidance. In: European Conference on Computer Vision. pp. 257–275. Springer (2024)

    2024

  21. [28]

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

    Hu, X., Fu, C.W., Zhu, L., Heng, P.A.: Depth-attentional features for single-image rain removal. In: Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition. pp. 8022–8031 (2019)

    2019

  22. [29]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Hu, X., Jiang, L., Schiele, B.: Training vision transformers for semi-supervised se- mantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 4007–4017 (2024)

    2024

  23. [30]

    arXiv preprint arXiv:1802.07934 (2018)

    Hung, W.C., Tsai, Y.H., Liou, Y.T., Lin, Y.Y., Yang, M.H.: Adversarial learn- ing for semi-supervised semantic segmentation. arXiv preprint arXiv:1802.07934 (2018)

    2018 arXiv

  24. [31]

    In: 2019 IEEE/CVF International Conference on Computer Vision Work- shop (ICCVW)

    Ji, B., Jung, H., Yoon, J., Kim, K., et al.: Bin-wise temperature scaling (bts): improvement in confidence calibration performance through simple scaling tech- niques. In: 2019 IEEE/CVF International Conference on Computer Vision Work- shop (ICCVW). pp. 4190–4196. IEEE (2019)

    2019

  25. [32]

    Information Processing & Management58(5), 102680 (2021) Rethinking Semi-supervised Segmentation Beyond Accuracy 15

    Jin, G., Liu, C., Chen, X.: Adversarial network integrating dual attention and sparse representation for semi-supervised semantic segmentation. Information Processing & Management58(5), 102680 (2021) Rethinking Semi-supervised Segmentation Beyond Accuracy 15

    2021

  26. [33]

    Advances in Neural Information Processing Systems35, 2803– 2816 (2022)

    Jin, Y., Wang, J., Lin, D.: Semi-supervised semantic segmentation via gentle teaching assistant. Advances in Neural Information Processing Systems35, 2803– 2816 (2022)

    2022

  27. [34]

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

    Kamann, C., Rother, C.: Benchmarking the robustness of semantic segmentation models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 8828–8838 (2020)

    2020

  28. [35]

    International journal of computer vision129(2), 462–483 (2021)

    Kamann, C., Rother, C.: Benchmarking the robustness of semantic segmentation models with respect to common corruptions. International journal of computer vision129(2), 462–483 (2021)

    2021

  29. [36]

    In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XIII 16

    Ke, Z., Qiu, D., Li, K., Yan, Q., Lau, R.W.: Guided collaborative training for pixel- wise semi-supervised learning. In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XIII 16. pp. 429–445. Springer (2020)

    2020

  30. [37]

    Advances in neural information processing systems32 (2019)

    Kull, M., Perello Nieto, M., K¨ angsepp, M., Silva Filho, T., Song, H., Flach, P.: Beyond temperature scaling: Obtaining well-calibrated multi-class probabilities with dirichlet calibration. Advances in neural information processing systems32 (2019)

    2019

  31. [38]

    In: DAGM German Conference on Pattern Recognition

    Landgraf, S., Hillemann, M., Kapler, T., Ulrich, M.: Efficient multi-task uncertain- ties for joint semantic segmentation and monocular depth estimation. In: DAGM German Conference on Pattern Recognition. pp. 348–364. Springer (2024)

    2024

  32. [40]

    ISPRS Annals of the Photogrammetry, Re- mote Sensing and Spatial Information Sciences10, 129–136 (2024)

    Landgraf, S., Hillemann, M., Wursthorn, K., Ulrich, M.: Uncertainty-aware cross- entropy for semantic segmentation. ISPRS Annals of the Photogrammetry, Re- mote Sensing and Spatial Information Sciences10, 129–136 (2024)

    2024

  33. [41]

    arXiv preprint arXiv:2501.08188 (2025)

    Landgraf, S., Qin, R., Ulrich, M.: A critical synthesis of uncertainty quantifi- cation and foundation models in monocular depth estimation. arXiv preprint arXiv:2501.08188 (2025)

    2025 arXiv

  34. [42]

    PFG–Journal of Photogrammetry, Remote Sensing and Geoinformation Science92(2), 101–114 (2024)

    Landgraf, S., Wursthorn, K., Hillemann, M., Ulrich, M.: Dudes: Deep uncer- tainty distillation using ensembles for semantic segmentation. PFG–Journal of Photogrammetry, Remote Sensing and Geoinformation Science92(2), 101–114 (2024)

    2024

  35. [43]

    Neurocomputing338, 321–348 (2019)

    Lateef, F., Ruichek, Y.: Survey on semantic segmentation using deep learning techniques. Neurocomputing338, 321–348 (2019)

    2019

  36. [44]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, D., Yang, J., Kreis, K., Torralba, A., Fidler, S.: Semantic segmentation with generative models: Semi-supervised learning and strong out-of-domain general- ization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 8300–8311 (2021)

    2021

  37. [45]

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

    Li, P., Purkait, P., Ajanthan, T., Abdolshah, M., Garg, R., Husain, H., Xu, C., Gould, S., Ouyang, W., Van Den Hengel, A.: Semi-supervised semantic seg- mentation under label noise via diverse learning groups. In: Proceedings of the IEEE/CVF International Conference on Compute...

    2023

  38. [46]

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

    Li, S., He, Y., Zhang, W., Zhang, W., Tan, X., Han, J., Ding, E., Wang, J.: Cfcg: Semi-supervised semantic segmentation via cross-fusion and contour guid- ance supervision. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 16348–16358 (2023)

    2023

  39. [47]

    arXiv preprint arXiv:2308.09281 (2023) 16 S

    Li, Y., Wang, X., Yang, L., Feng, L., Zhang, W., Gao, Y.: Diverse cotraining makes strong semi-supervised segmentor. arXiv preprint arXiv:2308.09281 (2023) 16 S. Landgraf et al

    2023 arXiv

  40. [48]

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

    Liang, C., Wang, W., Miao, J., Yang, Y.: Logic-induced diagnostic reasoning for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 16197–16208 (2023)

    2023

  41. [49]

    In: Computer vision–ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceedings, part v 13

    Lin, T.Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Doll´ ar, P., Zitnick, C.L.: Microsoft coco: Common objects in context. In: Computer vision–ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceedings, part v 13. pp. 740–75...

    2014

  42. [50]

    arXiv preprint arXiv:2104.04465 (2021)

    Liu, S., Zhi, S., Johns, E., Davison, A.J.: Bootstrapping semantic segmentation with regional contrast. arXiv preprint arXiv:2104.04465 (2021)

    2021 arXiv

  43. [51]

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

    Liu, Y., Tian, Y., Chen, Y., Liu, F., Belagiannis, V., Carneiro, G.: Perturbed and strict mean teachers for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 4258–4267 (2022)

    2022

  44. [52]

    Loiseau, T., Vu, T.H., Chen, M., P´ erez, P., Cord, M.: Reliability in semantic segmentation: Can we use synthetic data? In: European Conference on Computer Vision. pp. 442–459. Springer (2024)

    2024

  45. [53]

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

    Long, J., Shelhamer, E., Darrell, T.: Fully convolutional networks for semantic segmentation. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3431–3440 (2015)

    2015

  46. [54]

    In: Proceedings of the IEEE/CVF International Confer- ence on Computer Vision

    Ma, J., Wang, C., Liu, Y., Lin, L., Li, G.: Enhanced soft label for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF International Confer- ence on Computer Vision. pp. 1185–1195 (2023)

    2023

  47. [55]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Mai, H., Sun, R., Zhang, T., Wu, F.: Rankmatch: Exploring the better consistency regularization for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 3391– 3401 (2024)

    2024

  48. [56]

    In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIX 16

    Mendel, R., De Souza, L.A., Rauber, D., Papa, J.P., Palm, C.: Semi-supervised segmentation based on error-correcting supervision. In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIX 16. pp. 141–157. Springer (2020)

    2020

  49. [57]

    IEEE transactions on pattern analysis and machine intelligence44(7), 3523–3542 (2021)

    Minaee, S., Boykov, Y., Porikli, F., Plaza, A., Kehtarnavaz, N., Terzopoulos, D.: Image segmentation using deep learning: A survey. IEEE transactions on pattern analysis and machine intelligence44(7), 3523–3542 (2021)

    2021

  50. [58]

    IEEE transactions on pattern analysis and machine intelligence43(4), 1369–1379 (2019)

    Mittal, S., Tatarchenko, M., Brox, T.: Semi-supervised semantic segmentation with high-and low-level consistency. IEEE transactions on pattern analysis and machine intelligence43(4), 1369–1379 (2019)

    2019

  51. [59]

    IEEE Transactions on Intelligent Transportation Systems22(7), 4316–4336 (2020)

    Muhammad, K., Ullah, A., Lloret, J., Del Ser, J., De Albuquerque, V.H.C.: Deep learning for safe autonomous driving: Current challenges and future direc- tions. IEEE Transactions on Intelligent Transportation Systems22(7), 4316–4336 (2020)

    2020

  52. [60]

    arXiv preprint arXiv:1811.12709 (2018)

    Mukhoti, J., Gal, Y.: Evaluating bayesian deep learning methods for semantic segmentation. arXiv preprint arXiv:1811.12709 (2018)

    2018 arXiv

  53. [61]

    Advances in Neural Information Processing Systems36(2024)

    Na, J., Ha, J.W., Chang, H.J., Han, D., Hwang, W.: Switching temporary teach- ers for semi-supervised semantic segmentation. Advances in Neural Information Processing Systems36(2024)

    2024

  54. [62]

    In: Proceedings of the AAAI conference on artificial intelligence

    Naeini, M.P., Cooper, G., Hauskrecht, M.: Obtaining well calibrated probabili- ties using bayesian binning. In: Proceedings of the AAAI conference on artificial intelligence. vol. 29 (2015)

    2015

  55. [63]

    The International Journal of Robotics Research 37(4-5), 472–491 (2018) Rethinking Semi-supervised Segmentation Beyond Accuracy 17

    Oliveira, G.L., Bollen, C., Burgard, W., Brox, T.: Efficient and robust deep net- works for semantic segmentation. The International Journal of Robotics Research 37(4-5), 472–491 (2018) Rethinking Semi-supervised Segmentation Beyond Accuracy 17

    2018

  56. [64]

    In: Proceedings of the IEEE/CVF winter conference on applications of computer vision

    Olsson, V., Tranheden, W., Pinto, J., Svensson, L.: Classmix: Segmentation- based data augmentation for semi-supervised learning. In: Proceedings of the IEEE/CVF winter conference on applications of computer vision. pp. 1369–1378 (2021)

    2021

  57. [65]

    Transactions on Machine Learning Research Journal pp

    Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., et al.: Dinov2: Learning robust visual features without supervision. Transactions on Machine Learning Research Journal pp. 1–31 (2024)

    2024

  58. [66]

    Advances in neural information processing systems32(2019)

    Ovadia, Y., Fertig, E., Ren, J., Nado, Z., Sculley, D., Nowozin, S., Dillon, J., Lakshminarayanan, B., Snoek, J.: Can you trust your model’s uncertainty? eval- uating predictive uncertainty under dataset shift. Advances in neural information processing systems32(2019)

    2019

  59. [67]

    In: Proceedings of the IEEE international conference on computer vision

    Papandreou, G., Chen, L.C., Murphy, K.P., Yuille, A.L.: Weakly-and semi- supervised learning of a deep convolutional network for semantic image segmen- tation. In: Proceedings of the IEEE international conference on computer vision. pp. 1742–1750 (2015)

    2015

  60. [68]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Patra, R., Hebbalaguppe, R., Dash, T., Shroff, G., Vig, L.: Calibrating deep neural networks using explicit regularisation and dynamic data pruning. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 1541–1549 (2023)

    2023

  61. [69]

    Interna- tional Journal of Machine Learning and Cybernetics13(4), 1065–1077 (2022)

    Pedraza, A., Deniz, O., Bueno, G.: Really natural adversarial examples. Interna- tional Journal of Machine Learning and Cybernetics13(4), 1065–1077 (2022)

    2022

  62. [70]

    arXiv preprint arXiv:2302.09899 (2023)

    Pel´ aez-Vegas, A., Mesejo, P., Luengo, J.: A survey on semi-supervised semantic segmentation. arXiv preprint arXiv:2302.09899 (2023)

    2023 arXiv

  63. [71]

    Pattern Recognition107, 107269 (2020)

    Peng, J., Estrada, G., Pedersoli, M., Desrosiers, C.: Deep co-training for semi- supervised image segmentation. Pattern Recognition107, 107269 (2020)

    2020

  64. [72]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Qiao, P., Wei, Z., Wang, Y., Wang, Z., Song, G., Xu, F., Ji, X., Liu, C., Chen, J.: Fuzzy positive learning for semi-supervised semantic segmentation. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 15465–15474 (2023)

    2023

  65. [73]

    Recht, B., Roelofs, R., Schmidt, L., Shankar, V.: Do imagenet classifiers generalize to imagenet? In: International conference on machine learning. pp. 5389–5400. PMLR (2019)

    2019

  66. [74]

    International Journal of Computer Vision126, 973–992 (2018)

    Sakaridis, C., Dai, D., Van Gool, L.: Semantic foggy scene understanding with synthetic data. International Journal of Computer Vision126, 973–992 (2018)

    2018

  67. [75]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Sakaridis, C., Dai, D., Van Gool, L.: Acdc: The adverse conditions dataset with correspondences for semantic driving scene understanding. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 10765–10775 (2021)

    2021

  68. [76]

    The Bell system tech- nical journal27(3), 379–423 (1948)

    Shannon, C.E.: A mathematical theory of communication. The Bell system tech- nical journal27(3), 379–423 (1948)

    1948

  69. [77]

    Advances in neural information processing systems 33, 596–608 (2020)

    Sohn, K., Berthelot, D., Carlini, N., Zhang, Z., Zhang, H., Raffel, C.A., Cubuk, E.D., Kurakin, A., Li, C.L.: Fixmatch: Simplifying semi-supervised learning with consistency and confidence. Advances in neural information processing systems 33, 596–608 (2020)

    2020

  70. [78]

    In: Proceedings of the IEEE international conference on computer vision

    Souly, N., Spampinato, C., Shah, M.: Semi supervised semantic segmentation using generative adversarial network. In: Proceedings of the IEEE international conference on computer vision. pp. 5688–5696 (2017)

    2017

  71. [79]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition

    Sun, B., Yang, Y., Zhang, L., Cheng, M.M., Hou, Q.: Corrmatch: Label propaga- tion via correlation matching for semi-supervised semantic segmentation. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition. pp. 3097–3107 (2024) 18 S. Landgraf et al

    2024

  72. [80]

    Advances in neural information processing systems30(2017)

    Tarvainen, A., Valpola, H.: Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. Advances in neural information processing systems30(2017)

    2017

  73. [81]

    In: 2022 19th Conference on Robots and Vision (CR V)

    Teh, E.W., DeVries, T., Duke, B., Jiang, R., Aarabi, P., Taylor, G.W.: The gist and rist of iterative self-training for semi-supervised segmentation. In: 2022 19th Conference on Robots and Vision (CR V). pp. 58–66. IEEE (2022)

    2022

  74. [82]

    In: 2019 IEEE winter conference on applications of computer vision (W ACV)

    Varma, G., Subramanian, A., Namboodiri, A., Chandraker, M., Jawahar, C.: Idd: A dataset for exploring problems of autonomous navigation in unconstrained en- vironments. In: 2019 IEEE winter conference on applications of computer vision (W ACV). pp. 1743–1751. IEEE (2019)

    2019

  75. [83]

    In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision

    Wang, C., Xie, H., Yuan, Y., Fu, C., Yue, X.: Space engage: Collaborative space supervision for contrastive-based semi-supervised semantic segmentation. In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision. pp. 931–942 (2023)

    2023

  76. [84]

    Advances in Neural Information Pro- cessing Systems34, 11809–11820 (2021)

    Wang, D.B., Feng, L., Zhang, M.L.: Rethinking calibration of deep neural net- works: Do not be afraid of overconfidence. Advances in Neural Information Pro- cessing Systems34, 11809–11820 (2021)

    2021

  77. [85]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, H., Zhang, Q., Li, Y., Li, X.: Allspark: Reborn labeled features from unla- beled in transformer for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 3627–3636 (2024)

    2024

  78. [86]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, X., Bai, H., Yu, L., Zhao, Y., Xiao, J.: Towards the uncharted: Density- descending feature perturbation for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 3303–3312 (2024)

    2024

  79. [87]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, X., Zhang, B., Yu, L., Xiao, J.: Hunting sparsity: Density-guided con- trastive learning for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 3114–3123 (2023)

    2023

  80. [88]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition

    Wang, Y., Wang, H., Shen, Y., Fei, J., Li, W., Jin, G., Wu, L., Zhao, R., Le, X.: Semi-supervised semantic segmentation using unreliable pseudo-labels. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition. pp. 4248–4257 (2022)

    2022

  81. [89]

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

    Wang, Z., Zhao, Z., Xing, X., Xu, D., Kong, X., Zhou, L.: Conflict-based cross- view consistency for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 19585–19595 (2023)

    2023

  82. [90]

    Advances in Neural Information Processing Systems33, 4697– 4708 (2020)

    Wilson, A.G., Izmailov, P.: Bayesian deep learning and a probabilistic perspective of generalization. Advances in Neural Information Processing Systems33, 4697– 4708 (2020)

    2020

  83. [91]

    In: DAGM German Conference on Pattern Recognition

    Wolf, D.W., Balaji, P., Braun, A., Ulrich, M.: Decoupling of neural network cal- ibration measures. In: DAGM German Conference on Pattern Recognition. pp. 117–130. Springer (2024)

    2024

  84. [92]

    Multimedia Systems29(2), 511–523 (2023)

    Wu, Y., Liu, C., Chen, L., Zhao, D., Zheng, Q., Zhou, H.: Perturbation consis- tency and mutual information regularization for semi-supervised semantic seg- mentation. Multimedia Systems29(2), 511–523 (2023)

    2023

  85. [93]

    Ad- vances in neural information processing systems34, 12077–12090 (2021) Rethinking Semi-supervised Segmentation Beyond Accuracy 19

    Xie, E., Wang, W., Yu, Z., Anandkumar, A., Alvarez, J.M., Luo, P.: Segformer: Simple and efficient design for semantic segmentation with transformers. Ad- vances in neural information processing systems34, 12077–12090 (2021) Rethinking Semi-supervised Segmentation Beyond Accuracy 19

    2021

  86. [94]

    Advances in neural information pro- cessing systems35, 26007–26020 (2022)

    Xu, H., Liu, L., Bian, Q., Yang, Z.: Semi-supervised semantic segmentation with prototype-based consistency regularization. Advances in neural information pro- cessing systems35, 26007–26020 (2022)

    2022

  87. [95]

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

    Yang, L., Qi, L., Feng, L., Zhang, W., Shi, Y.: Revisiting weak-to-strong consistency in semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 7236– 7246 (2023)

    2023

  88. [96]

    IEEE Transactions on Pattern Analysis and Machine In- telligence (2025)

    Yang, L., Zhao, Z., Zhao, H.: Unimatch v2: Pushing the limit of semi-supervised semantic segmentation. IEEE Transactions on Pattern Analysis and Machine In- telligence (2025)

    2025

  89. [97]

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

    Yang, L., Zhuo, W., Qi, L., Shi, Y., Gao, Y.: St++: Make self-training work better for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 4268–4277 (2022)

    2022

  90. [98]

    IEEE transactions on knowledge and data engineering35(9), 8934–8954 (2022)

    Yang, X., Song, Z., King, I., Xu, Z.: A survey on deep semi-supervised learning. IEEE transactions on knowledge and data engineering35(9), 8934–8954 (2022)

    2022

  91. [99]

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

    Yu, X., Zuo, Y., Wang, Z., Zhang, X., Zhao, J., Yang, Y., Jiao, L., Peng, R., Wang, X., Zhang, J., et al.: The robust semantic segmentation uncv2023 challenge results. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 4618–4628 (2023)

    2023

  92. [100]

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

    Yuan, J., Liu, Y., Shen, C., Wang, Z., Li, H.: A simple baseline for semi-supervised semantic segmentation with strong data augmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 8229–8238 (2021)

    2021

  93. [101]

    Journal of Visual Communication and Image Representation78, 103170 (2021)

    Zhang, J., Li, Z., Zhang, C., Ma, H.: Stable self-attention adversarial learning for semi-supervised semantic image segmentation. Journal of Visual Communication and Image Representation78, 103170 (2021)

    2021

  94. [102]

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

    Zhao, H., Shi, J., Qi, X., Wang, X., Jia, J.: Pyramid scene parsing network. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 2881–2890 (2017)

    2017

  95. [103]

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

    Zhao, Z., Long, S., Pi, J., Wang, J., Zhou, L.: Instance-specific and model-adaptive supervision for semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 23705– 23714 (2023)

    2023

  96. [104]

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

    Zhao, Z., Yang, L., Long, S., Pi, J., Zhou, L., Wang, J.: Augmentation mat- ters: A simple-yet-effective approach to semi-supervised semantic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 11350–11359 (2023)

    2023

  97. [105]

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

    Zhou, B., Zhao, H., Puig, X., Fidler, S., Barriuso, A., Torralba, A.: Scene parsing through ade20k dataset. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 633–641 (2017)

    2017

  98. [106]

    In: International conference on machine learning

    Zhou, D., Yu, Z., Xie, E., Xiao, C., Anandkumar, A., Feng, J., Alvarez, J.M.: Understanding the robustness in vision transformers. In: International conference on machine learning. pp. 27378–27394. PMLR (2022)

    2022

  99. [107]

    IEEE Transactions on Intelli- gent Transportation Systems21(5), 1951–1963 (2019)

    Zhou, W., Berrio, J.S., Worrall, S., Nebot, E.: Automated evaluation of semantic segmentation robustness for autonomous driving. IEEE Transactions on Intelli- gent Transportation Systems21(5), 1951–1963 (2019)

    2019

  100. [108]

    IEEE Access10, 48855–48864 (2022)

    Zhou, Y., Jiao, R., Wang, D., Mu, J., Li, J.: Catastrophic forgetting problem in semi-supervised semantic segmentation. IEEE Access10, 48855–48864 (2022)

    2022

  101. [109]

    IEEE transactions on pattern analysis and machine intelligence46(3), 1589–1602 (2021) 20 S

    Zhu, Y., Zhang, Z., Wu, C., Zhang, Z., He, T., Zhang, H., Manmatha, R., Li, M., Smola, A.: Improving semantic segmentation via efficient self-training. IEEE transactions on pattern analysis and machine intelligence46(3), 1589–1602 (2021) 20 S. Landgraf et al. Supplementary Mat...

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.