Pith. sign in

REVIEW 5 major objections 5 minor 49 references

FARCLUSS: Fuzzy Adaptive Rebalancing and Contrastive Uncertainty Learning for Semi-Supervised Semantic Segmentation

T0 review · 5 major / 5 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 can learn from uncertain pixels by keeping soft top-K labels and entropy-based weights, rather than discarding ambiguous predictions.

desk verdict A plausible combination of known semi-supervised segmentation ideas whose own tables contradict the 'outperforms SOTA' claim; the ablations are solid but the paper needs major claim-softening and a direct comparison to its own fuzzy-labeling source. read the letter →

arxiv 2506.11142 v2 pith:SUEZZVJB submitted 2025-06-11 cs.CV cs.LGeess.IV

Ebenezer Tarubinga , Jenifer Kalafatovich , Seong-Whan Lee This is my paper · ORCID
classification cs.CVcs.LGeess.IV
keywords semi-supervisedlearningsemanticsegmentationfuzzypseudo-labelinguncertaintyestimationclassimbalancecontrastivemeanteacherentropyweighting
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

This paper tries to establish that semi-supervised semantic segmentation performs better when a model uses, rather than discards, its uncertain predictions. It proposes FARCLUSS, a teacher–student framework whose unsupervised loss is built from fuzzy pseudo-labels: for each pixel it keeps the top-K class probabilities instead of a single hard label. Those soft labels are weighted pixel-wise by a reliability score derived from prediction entropy, reweighted per class by the inverse of that class's frequency in the current batch, and supplemented by a cheap prototype-based contrastive loss that pulls pixel embeddings toward class centroids. On Pascal VOC 2012 and Cityscapes, the authors report consistent mIoU improvements over comparison methods, with the largest advantages at low label ratios (1/16 and 1/8).

What carries the argument

The load-bearing object is the fuzzy soft pseudo-label: a per-pixel distribution formed by taking the teacher's top-K class probabilities and renormalizing them, so each pixel keeps a small set of plausible classes instead of being forced into one hard label. The argument runs on three additional mechanisms attached to that label: an entropy-derived pixel weight $W_{h,w}=1-H(p^T_{:,h,w})$ that continuously down-weights ambiguous locations, a batch-level class weight $w_c=\mathrm{median}(F)/(F_c+\epsilon)$ that boosts rare classes, and a prototype contrastive loss that averages high-confidence embedded pixels per class and penalizes cosine distance from those centroids. Together they let the unsupervised loss carry soft, reliability-weighted, rebalanced supervision into the student model.

What would settle it

Re-run FARCLUSS and its nearest baselines (especially UniMatch and PS-MT) under one shared codebase with identical augmentations, crop sizes, training lengths, and at least three random seeds, at the 1/16 Pascal VOC split; if the reported mIoU advantage of about 1 point does not reproduce or reverses, the central superiority claim fails.

Watch

Extended reading notes

Core claim

The central discovery claimed is that uncertainty in teacher predictions is a reusable resource rather than noise to be filtered out. The paper's recipe is to transform each teacher probability map into a fuzzy label by renormalizing the probabilities of the top-K classes per pixel, then treat the remaining entropy as a continuous confidence signal. A per-pixel weight $W_{h,w}=1-H(p^T_{:,h,w})$ down-weights high-entropy pixels, a per-class weight $w_c=\mathrm{median}(F)/(F_c+\epsilon)$ amplifies rare classes in each batch, and a lightweight contrastive term on class prototypes regularizes feature geometry. The authors argue—and support with ablations—that the combination yields super-additive gains: removing fuzzy labeling costs about 6.2 mIoU on Pascal VOC 1/8, entropy weighting about 3.5, class rebalancing 2.6 on Cityscapes, and contrastive regularization 0.5–0.7.

Load-bearing premise

The reported gains over prior methods are small—often below 1 mIoU point—and the comparison assumes identical training and evaluation settings across papers; if those settings actually differ, the ranking could flip.

Editorial extensions

If this is right

  • If the claimed gains hold, then strict confidence thresholding—the default in many FixMatch-style pipelines—is leaving useful signal on the table, and soft top-K labels can recover part of it.
  • At low label ratios (1/16 and 1/8), the method's reported edge over its closest baselines is largest, suggesting uncertainty-preserving supervision matters most when labeled data are scarce.
  • The per-batch class rebalancing scheme implies that class-imbalance compensation can be done locally and adaptively, without needing global class priors or manual tuning.
  • The prototype-based contrastive loss offers a low-overhead way to obtain feature compactness, pointing to a design where contrastive regularization is affordable in single-network teacher–student settings.
  • Because all components plug into a single shared teacher–student architecture, the method keeps the data requirements of streamlined methods like UniMatch while adding four corrective terms.

Reading between the lines

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

  • The fuzzy pseudo-label can be read as a form of label smoothing that is adaptive per pixel, which might explain both the stability and the boundary-region improvements; one could test this by ablating top-K size K and comparing against fixed label smoothing.
  • The entropy weight $W_{h,w}=1-H$ acts as a soft curriculum that naturally up-weights easy pixels early and gradually admits harder ones, so a separate curriculum schedule may be unnecessary.
  • Per-batch inverse-frequency reweighting can be unstable in very small batches; a natural extension is to mix batch statistics with an exponentially averaged class-frequency estimate, which the paper does not explore.
  • The same four-component recipe could transfer to other dense prediction tasks, such as medical image segmentation, where boundary ambiguity and class imbalance are similarly dominant failure modes.
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. FARCLUSS is a semi-supervised semantic segmentation method built on a mean-teacher architecture. It proposes four components: (i) fuzzy soft pseudo-labels obtained by normalizing the teacher's top-K class probabilities; (ii) pixel-wise weights based on normalized entropy; (iii) batch-level adaptive class rebalancing using median-frequency inverse weights; and (iv) a prototype-based contrastive loss on high-confidence pixels. The paper reports mIoU comparisons on PASCAL VOC (classic and blended) and Cityscapes with ResNet-50/101 across 1/16, 1/8, 1/4, and 1/2 labeled splits, plus ablations and class-level analyses. The central claim is that FARCLUSS outperforms current state-of-the-art approaches, especially for under-represented classes and ambiguous regions.

Significance. If the empirical claims were fully supported, the framework would be a useful contribution: it targets two known weaknesses of semi-supervised segmentation (discarded uncertain pseudo-labels and class imbalance), and the individual loss terms in Eqs. (2)-(11) are coherent and computationally inexpensive. The ablation in Table 5 indicates that each component contributes, with fuzzy labeling having the largest effect. However, the significance is reduced by the fact that the headline state-of-the-art claim is not consistently supported by the paper's own tables: in several settings the method trails strong published baselines, most reported differences are below 1 mIoU point, and no variance estimates are provided. In addition, Algorithm 1 contains tensor-shape and batching errors that prevent reproduction. The paper's value is currently an incremental combination of existing ideas rather than a clearly demonstrated new state of the art.

major comments (5)
  1. [Abstract; Tables 1-3] The abstract's claim that the method 'outperforms current state-of-the-art approaches' is contradicted by the paper's own comparisons. In Table 1 (Pascal classic, ResNet-101), FARCLUSS scores 76.4/78.2/79.0/80.3 at 1/16, 1/8, 1/4, and 1/2, while CorrMatch scores 76.4/78.5/79.4/80.6, Diverse CoT reaches 80.1 at 1/4, and DDFP reaches 81.2 at 1/2. In Table 2 (Blended, train size 513, ResNet-101), CorrMatch is higher at 1/8 (79.3 vs. 79.0). In Table 3 (Cityscapes, ResNet-101), LogicDiag is higher at 1/8 (78.9 vs. 78.5) and 1/4 (80.2 vs. 80.0) and ties at 1/2 (81.0). Since the central contribution rests on empirical superiority, the authors should either soften the claim to 'competitive' or provide additional matched-protocol evidence that the method is actually ahead.
  2. [Algorithm 1] Algorithm 1 is not runnable as written. The teacher output `pw` is produced from `xw` with batch size B, so `fuzzy` and `w_px` have batch dimension B, while `feat = g(torch.cat([xs1, xs2]))` has batch dimension 2B; consequently `criterion_ce(pred.log_softmax(1), fuzzy)` and the multiplication `w_px.view(-1,1,1,1) * w_cls[lbl].view(-1,1,1,1) * loss_u_map` have incompatible leading dimensions. The selection `sel = w_px > 0.5` cannot index `feat` for the same reason. The channel-masking step `pred = h(feat * mask)` is not described in Section 3, and `nn.CrossEntropyLoss()(f(x_l).log_softmax(1), y_l)` applies a log-softmax before a loss that already expects raw logits. Please correct these inconsistencies and state precisely how the weak view and the two strong views are batched.
  3. [Eq. (7); Algorithm 1] The adaptive class rebalancing uses the median of the raw per-batch class counts `cnt`, including classes that do not occur in the batch. In small batches (e.g., Cityscapes with a 19-class label space), the median will frequently be zero, so `w_cls = 0/(cnt+1e-6)` becomes zero for every class that actually appears, driving the unsupervised loss to zero; when the median is positive, the weights remain highly sensitive to which classes were sampled. Please define the statistic over non-zero frequencies or use a smoothed count, and report the batch size and sensitivity to this choice.
  4. [Tables 1-3] Most reported differences between FARCLUSS and the closest baselines are below 1 mIoU point (for example, Table 1, ResNet-50 1/16: 72.90 vs. 72.80 for CW-BASS; Table 3, ResNet-101 1/16: 77.2 vs. 77.3 for CorrMatch, where FARCLUSS is actually lower). No standard deviations, seeds, or reimplementations of baselines are reported, so the claimed advantage is not statistically distinguishable from noise. The authors should add variance estimates across multiple seeds and, where feasible, re-run the strongest baselines under the same training protocol.
  5. [Eq. (8) vs. Algorithm 1] The class-rebalancing implementation in Algorithm 1 does not match Eq. (8). Equation (8) multiplies each class term inside the KL divergence by `w_c`, whereas the algorithm computes `w_cls[argmax(fuzzy)]` and multiplies the entire per-pixel KL sum by this single weight. For a pixel whose fuzzy label distributes mass over two classes, which is the intended behavior of the top-K fuzzy labeling, the two formulations are quantitatively different. Please align the algorithm with Eq. (8) or explicitly define the argmax-based approximation and justify it.
minor comments (5)
  1. [Section 4.2.4] The implementation details state a normalized entropy threshold of 0.7, but the loss definitions in Eqs. (4)-(8) and Algorithm 1 do not refer to this threshold; please clarify where and how the threshold is applied.
  2. [Figure 3] The caption says all methods are trained under the Full (1464) setting, while the text discusses semi-supervised settings; please specify the exact labeled fraction and training protocol for each qualitative comparison.
  3. [Table 6] The class-level comparison introduces CPCL and DUEB without stating the training resolution, number of labeled images, or augmentation protocol for these runs; please add this information so the per-class numbers are interpretable.
  4. [Section 4.5] The text claims a 'monotonic descent' of the total loss and attributes spikes to bursts in L_u, but the figure lacks axis labels and the claims are not quantified; please add a properly labeled figure and describe whether the curves are smoothed.
  5. [Throughout] The text contains typos and notation glitches: 'that modulate' should be 'that modulates', several Greek symbols appear as 'Îż' in Algorithm 1, and venue names are inconsistent ('IJCNN' vs. 'IJCNN').

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the reported results rest on external benchmark comparisons and standard consistency losses, with no equation reducing to its own input.

full rationale

The paper's central claim is an empirical superiority result on PASCAL VOC and Cityscapes, judged against published mIoU numbers from external baselines. The loss terms in Eqs. (4), (8), and (10) are respectively a KL consistency term, a reweighted KL term, and a prototype-variance term; none of them defines the reported validation mIoU in terms of the fitted weights or pseudo-labels. Fuzzy pseudo-labeling in Eqs. (2)-(3) is explicitly attributed to external work via citation [47], so while the contribution list may overstate novelty, that is an attribution issue rather than a circular derivation. The only same-author reference, CW-BASS [5], appears as a comparison baseline and as an efficiency reference in Table 4; it is not load-bearing for the method's design or for any reported result. No parameter is fitted to a target and then renamed as a prediction, and no uniqueness claim is imported from the authors' prior work. The mismatch between the abstract's 'outperforms' wording and some table entries is a correctness or consistency concern, not a circularity concern.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical; no new physical or mathematical entities are introduced. It depends on standard mean-teacher and uncertainty-calibration assumptions, on hand-set hyperparameters, and on treating published benchmark numbers as directly comparable. The fuzzy-label component is borrowed from [47], so no new mechanism is postulated.

free parameters (6)
  • K (top-K fuzzy labels) = 2
    Set to 2 for both datasets in Section 4.2.4; controls how much class ambiguity is preserved.
  • lambda_u (unsupervised loss weight) = 0.5
    Set in Section 3.7.1; balances supervised and unsupervised terms.
  • lambda_c (contrastive loss weight) = 0.1
    Chosen based on preliminary experiments in Section 4.2.4; affects how strongly embeddings are pulled to prototypes.
  • Entropy threshold = 0.7
    Defines high-uncertainty pixels in Section 4.2.4; interacts with pixel weighting.
  • EMA momentum alpha = 0.99
    Fixed mean-teacher momentum in Eq. (1); standard value, but affects pseudo-label stability.
  • Projection head dimension = 128
    Dimensionality of contrastive embeddings in Section 4.2.4; design choice.
assumptions (5)
  • domain assumption EMA teacher predictions are stable enough to serve as pseudo-labels.
    Standard mean-teacher assumption; Eq. (1) sets alpha=0.99.
  • domain assumption Normalized entropy H(p_T) is a valid per-pixel reliability score.
    Eqs. (5)-(6) assume high entropy correlates with label noise.
  • ad hoc to paper Per-batch median-frequency reweighting improves minority-class learning without destabilizing training.
    Eq. (7) is a heuristic; no analysis of batch composition effects.
  • domain assumption Cross-paper benchmark numbers in Tables 1-3 are directly comparable.
    The headline claim depends on identical training and evaluation protocols across papers.
  • domain assumption Class-mean prototypes are sufficient summaries for contrastive regularization.
    Eqs. (9)-(10) assume the mean embedding captures class structure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FARCLUSS: Fuzzy Adaptive Rebalancing and Contrastive Uncertainty Learning for Semi-Supervised Semantic Segmentation." pith.science (2026). https://pith.science/paper/SUEZZVJB

@misc{pith2026250611142,
  author       = {Pith},
  title        = {Pith review of: FARCLUSS: Fuzzy Adaptive Rebalancing and Contrastive Uncertainty Learning for Semi-Supervised Semantic Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SUEZZVJB}},
  note         = {Machine review of arXiv:2506.11142}
}
read the original abstract

Semi-supervised semantic segmentation (SSSS) faces persistent challenges in effectively leveraging unlabeled data, such as ineffective utilization of pseudo-labels, exacerbation of class imbalance biases, and neglect of prediction uncertainty. Current approaches often discard uncertain regions through strict thresholding favouring dominant classes. To address these limitations, we introduce a holistic framework that transforms uncertainty into a learning asset through four principal components: (1) fuzzy pseudo-labeling, which preserves soft class distributions from top-K predictions to enrich supervision; (2) uncertainty-aware dynamic weighting, that modulate pixel-wise contributions via entropy-based reliability scores; (3) adaptive class rebalancing, which dynamically adjust losses to counteract long-tailed class distributions; and (4) lightweight contrastive regularization, that encourage compact and discriminative feature embeddings. Extensive experiments on benchmarks demonstrate that our method outperforms current state-of-the-art approaches, achieving significant improvements in the segmentation of under-represented classes and ambiguous regions.

Figures

Figures reproduced from arXiv: 2506.11142 by the authors.

Figure 1
Figure 1. Pixel-wise normalized entropy map from the teacher model on PASCAL VOC samples. Brighter colors represent higher prediction uncertainty. Uncertainty is concentrated at object boundaries and cluttered background regions. the batch level. In contrast, our adaptive class rebalancing technique leverages per￾batch pseudo-label frequencies to scale losses, ensuring that classes like traffic sign in Cityscapes receive incr… view at source ↗
Figure 2
Figure 2. Overview of the FARCLUSS Framework. FARCLUSS employs a teacher-student framework where labeled data supervise the student directly, while unlabeled data undergo fuzzy pseudo-labeling and uncertainty-based weighting via the teacher. Adaptive class rebalancing and prototype-based contrastive learning further refine the total loss. 3. Method Our method integrates fuzzy pseudo-labeling, uncertainty-based dynamic weight￾… view at source ↗
Figure 3
Figure 3. Qualitative results of our method, FARCLUSS with other state-of-the-art methods on the PASCAL VOC 2012 dataset. All methods are trained under the Full (1464) setting for validation. Sup Only refers to supervised baseline trained only on the corresponding proportion of labeled data. Red rectangles highlight regions where segmentation performance is improved. (1/8 labeled data) datasets using a ResNet-101 backbone. We… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results of our method, FARCLUSS with other state-of-the-art methods on the Cityscapes dataset. All methods are trained under the 1/8 (12.5%) setting for validation. Red rectangles highlight regions where segmentation performance is improved. 4.4.1. Componen…
Figure 5
Figure 5. Figure 5: Training trajectories of the total loss Ltotal alongside its supervised component, L𝑠 and unsupervised component, L𝑢 over 200 k iterations on Pascal VOC with a 1/16 labelled split. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 45 canonical work pages

  1. [47]

    Fuzzy positive learning for semi-supervised semantic segmentation,

    P. Qiao, Z. Wei, Y. Wang, Z. Wang, G. Song, F. Xu, X. Ji, C. Liu, and J. Chen, “Fuzzy positive learning for semi-supervised semantic segmentation,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2023, pp. 15465–15474

    work page 2023

  2. [1]

    ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation,

    T.-H. Vu, H. Jain, M. Bucher, M. Cord, and P. PÃľrez, “ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation,” in CVPR, pp. 2517–2526, 2019

    work page 2019

  3. [2]

    Perturbed and strict mean teachers for semi-supervised semantic segmentation,

    Y. Liuet al., “Perturbed and strict mean teachers for semi-supervised semantic segmentation,” inCVPR, pp. 4258–4267, 2022

    work page 2022

  4. [3]

    Revisiting Weak-to-Strong Consistency in Semi-Supervised Se- mantic Segmentation,

    L. Yanget al., “Revisiting Weak-to-Strong Consistency in Semi-Supervised Se- mantic Segmentation,” inCVPR, 2023

    work page 2023

  5. [4]

    Reco: Retrieve and co-segment for zero-shot transfer,

    G. Shin, W. Xie, and S. Albanie, “Reco: Retrieve and co-segment for zero-shot transfer,”NeurIPS, vol. 35, pp. 33754–33767, 2022

    work page 2022

  6. [5]

    CW-BASS: Confidence- Weighted Boundary-Aware Learning for Semi-Supervised Semantic Segmenta- tion,

    E. Tarubinga, J. Kalafatovich and Seong-Whan Lee “CW-BASS: Confidence- Weighted Boundary-Aware Learning for Semi-Supervised Semantic Segmenta- tion,”arXiv preprint arXiv:2502.15152, 2025

    arXiv 2025

  7. [6]

    Semisupervised semantic segmentation with cross pseudo super- vision,

    X. Chenet al., “Semisupervised semantic segmentation with cross pseudo super- vision,” inCVPR, pp. 2613–2622, 2021

    work page 2021

  8. [7]

    Fixmatch: Simplifyingsemi-supervisedlearningwithconsistency and confidence,

    K.Sohn etal.,“Fixmatch: Simplifyingsemi-supervisedlearningwithconsistency and confidence,” inNeurIPS, 2020

    work page 2020

Show all 49 references
  1. [8]

    Pseudoseg: Designing pseudo labels for semantic segmentation,

    Y. Zouet al., “Pseudoseg: Designing pseudo labels for semantic segmentation,” arXiv preprint arXiv:2010.09713, 2020

    2010 arXiv

  2. [9]

    Semi-supervised semantic segmentation with directional context- aware consistency,

    X. Laiet al., “Semi-supervised semantic segmentation with directional context- aware consistency,” inCVPR, pp. 1205–1214, 2021

    2021

  3. [10]

    Semi-supervised semantic segmentation with error lo- calization network,

    D. Kwon and S. Kwak, “Semi-supervised semantic segmentation with error lo- calization network,” inCVPR, pp. 9957–9967, 2022

    2022

  4. [11]

    Semi-supervisedsemanticsegmentationusingunreliablepseudo- labels,

    Y.Wangetal.,“Semi-supervisedsemanticsegmentationusingunreliablepseudo- labels,” inCVPR, pp. 4248–4257, 2022. 26

    2022

  5. [12]

    Pseudo-Label: The Simple and Efficient Semi-Supervised Learn- ing Method for Deep Neural Networks,

    D. H. Lee, “Pseudo-Label: The Simple and Efficient Semi-Supervised Learn- ing Method for Deep Neural Networks,” inICML Workshop on Challenges in Representation Learning, pp. 896–901, 2013

    2013

  6. [13]

    Temporalensembling forsemi-supervisedlearning,

    S.Laine andT.Aila, “Temporalensembling forsemi-supervisedlearning,” arXiv preprint arXiv:1610.02242, 2016

    2016 arXiv

  7. [14]

    Deep residual learning for image recognition,

    K. Heet al., “Deep residual learning for image recognition,” inCVPR, pp. 770– 778, 2016

    2016

  8. [15]

    Imagenet: A large-scale hierarchical image database,

    J. Denget al., “Imagenet: A large-scale hierarchical image database,” inCVPR, pp. 248–255, 2009

    2009

  9. [16]

    RepUNet: A fast image semantic segmentation model based on convolutional reparameterization of ship satellite images,

    Y. Zhenget al., “RepUNet: A fast image semantic segmentation model based on convolutional reparameterization of ship satellite images,” inCACRE, pp. 461– 465, IEEE, 2021

    2021

  10. [17]

    Bayesian Nested Neural Networks for Uncertainty Calibration and Adaptive Compression,

    Y. Cuiet al., “Bayesian Nested Neural Networks for Uncertainty Calibration and Adaptive Compression,” inCVPR, pp. 2392–2401, 2021

    2021

  11. [18]

    DARS: Data Augmentation using Refined Segmentation on Computer Vision Tasks,

    Y. H. Choet al., “DARS: Data Augmentation using Refined Segmentation on Computer Vision Tasks,” inICTC, pp. 348–350, IEEE, 2021

    2021

  12. [19]

    Semi-supervised semantic segmentation via adaptive equalization learning,

    H. Huet al., “Semi-supervised semantic segmentation via adaptive equalization learning,”NeurIPS, vol. 34, pp. 22106–22118, 2021

    2021

  13. [20]

    UCC: Uncertainty Guided Cross-Head Co-Training for Semi- Supervised Semantic Segmentation,

    J. Fan et al., “UCC: Uncertainty Guided Cross-Head Co-Training for Semi- Supervised Semantic Segmentation,” inCVPR, pp. 9947–9956, 2022

    2022

  14. [21]

    Instance-Specific and Model-Adaptive Supervision for Semi- Supervised Semantic Segmentation,

    Z. Zhao et al., “Instance-Specific and Model-Adaptive Supervision for Semi- Supervised Semantic Segmentation,” inCVPR, pp. 23705–23714, 2023

    2023

  15. [22]

    Semi-Supervised Semantic Segmentation with Cross-Consistency Training,

    Y. Ouali, C. Hudelot, and M. Tami, “Semi-Supervised Semantic Segmentation with Cross-Consistency Training,” inCVPR, pp. 12674–12684, 2020

    2020

  16. [23]

    Multi-Granularity Distillation Scheme Towards Lightweight Semi- Supervised Semantic Segmentation,

    J. Qinet al., “Multi-Granularity Distillation Scheme Towards Lightweight Semi- Supervised Semantic Segmentation,” inECCV, pp. 481–498, Springer, 2022. 27

    2022

  17. [24]

    Mean teachers are better role models: Weight- averaged consistency targets improve semi-supervised deep learning results,

    A. Tarvainen and H. Valpola, “Mean teachers are better role models: Weight- averaged consistency targets improve semi-supervised deep learning results,” in NeurIPS, 2017

    2017

  18. [25]

    Deeplab: Semantic image segmentation with deep convolu- tional nets, atrous convolution, and fully connected crfs,

    L.-C. Chenet al., “Deeplab: Semantic image segmentation with deep convolu- tional nets, atrous convolution, and fully connected crfs,”TPAMI, vol. 40, no. 4, pp. 834–848, 2017

    2017

  19. [26]

    A simple framework for contrastive learning of visual representa- tions,

    T. Chenet al., “A simple framework for contrastive learning of visual representa- tions,” inICML, 2020

    2020

  20. [27]

    C3-SemiSeg: Contrastive Semi-supervised Segmentation via Cross-set Learning and Dynamic Class-Balancing,

    T. Zhou et al., “C3-SemiSeg: Contrastive Semi-supervised Segmentation via Cross-set Learning and Dynamic Class-Balancing,” inICCV, 2021

    2021

  21. [28]

    ST++: Makeself-trainingworkbetterforsemi-supervisedsemantic segmentation,

    J.Yangetal.,“ST++: Makeself-trainingworkbetterforsemi-supervisedsemantic segmentation,” inCVPR, 2022

    2022

  22. [29]

    Semi-supervisedsemanticsegmentationusingunreliablepseudo- labels,

    Y.Wangetal.,“Semi-supervisedsemanticsegmentationusingunreliablepseudo- labels,” inCVPR, 2022

    2022

  23. [30]

    Semi-supervised semantic segmentation with generalized task- aware consistency,

    X. Chenet al., “Semi-supervised semantic segmentation with generalized task- aware consistency,” inNeurIPS, 2022

    2022

  24. [31]

    Semi-supervised semantic segmentation with pixel contrastive consistency,

    T. Zhouet al., “Semi-supervised semantic segmentation with pixel contrastive consistency,” inNeurIPS, 2022

    2022

  25. [32]

    AugSeg: Asimpledataaugmentationmethodforsemi-supervised semantic segmentation,

    Y.Zhangetal.,“AugSeg: Asimpledataaugmentationmethodforsemi-supervised semantic segmentation,” inCVPR, 2023

    2023

  26. [33]

    Diverse co-training for semi-supervised semantic segmentation,

    Y. Liet al., “Diverse co-training for semi-supervised semantic segmentation,” in ICCV, 2023

    2023

  27. [34]

    Semi-supervisedsemanticsegmentationwithentropy-basedsam- ple learning,

    X.Chen etal.,“Semi-supervisedsemanticsegmentationwithentropy-basedsam- ple learning,” inICCV, 2023

    2023

  28. [35]

    Semi-supervised semantic segmentation with logical diagnosis,

    T. Zhouet al., “Semi-supervised semantic segmentation with logical diagnosis,” in ICCV, 2023. 28

    2023

  29. [36]

    Semi-supervised semantic segmentation with dual adaptive weight- ing,

    Y. Liet al., “Semi-supervised semantic segmentation with dual adaptive weight- ing,” inNeurIPS, 2023

    2023

  30. [37]

    Semi-supervisedsemanticsegmentationwithdynamicdualfeature propagation,

    X.Chen etal.,“Semi-supervisedsemanticsegmentationwithdynamicdualfeature propagation,” inCVPR, 2024

    2024

  31. [38]

    Semi-supervised semantic segmentation with correspondence matching,

    T. Zhou et al., “Semi-supervised semantic segmentation with correspondence matching,” inCVPR, 2024

    2024

  32. [39]

    UniMatch: Unified semi-supervised learning for classification and segmentation,

    Y. Zhanget al., “UniMatch: Unified semi-supervised learning for classification and segmentation,” inCVPR, 2023

    2023

  33. [40]

    The Pascal Visual Object Classes Challenge: A Retrospective,

    M. Everingham, S. A. Eslami, L. Van Gool, C. K. Williams, J. Winn, and A. Zisserman, “The Pascal Visual Object Classes Challenge: A Retrospective,”In- ternationalJournalofComputerVision(IJCV) ,vol.111,no.1,pp.98–136,2015

    2015

  34. [41]

    The Cityscapes Dataset for Semantic Urban Scene Understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benenson, U. Franke, S. Roth, and B. Schiele, “The Cityscapes Dataset for Semantic Urban Scene Understanding,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 3213–3223

    2016

  35. [42]

    MicrosoftCOCO:CommonObjectsinContext,

    T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. DollÃąr, andC.L.Zitnick,“MicrosoftCOCO:CommonObjectsinContext,”in European Conference on Computer Vision (ECCV), 2014, pp. 740–755

    2014

  36. [43]

    Uncertainty and Energy based Loss Guided Semi-Supervised Semantic Segmentation,

    Smita Thakur, Rini and Vinod K. Kurmi, “Uncertainty and Energy based Loss Guided Semi-Supervised Semantic Segmentation,”arXiv e-prints, arXiv:2501, 2025

    2025

  37. [44]

    Semi- supervised semantic segmentation needs strong, varied perturbations,

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

    1906 arXiv

  38. [45]

    Guidedcollaborativetrainingforpixel-wise semi-supervised learning,

    Z.Xia,Y.Liu,Y.Wei,andY.Yang,“Guidedcollaborativetrainingforpixel-wise semi-supervised learning,”arXiv preprint arXiv:2008.05258, 2020. 29

    2008 arXiv

  39. [46]

    Learn- ing from future: A novel self-training framework for semantic segmentation,

    Y.Du,Y.Shen,H.Wang,J.Fei,W.Li,L.Wu,R.Zhao,Z.Fu,andQ.Liu,“Learn- ing from future: A novel self-training framework for semantic segmentation,” in Advances in Neural Information Processing Systems, vol. 35, 2022

    2022

  40. [48]

    Semantic contours from inverse detectors,

    B. Hariharan, P. Arbeláez, L. Bourdev, S. Maji, and J. Malik, “Semantic contours from inverse detectors,” inProceedings of the 2011 International Conference on Computer Vision, IEEE, 2011, pp. 991–998

    2011

  41. [49]

    Conservative-progressive collaborative learning for semi-supervised semantic segmentation,

    S. Fan, F. Zhu, Z. Feng, Y. Lv, M. Song, and F. Wang, “Conservative-progressive collaborative learning for semi-supervised semantic segmentation,”IEEE Trans- actions on Image Processing, vol. 32, pp. 6183–6194, 2023. 30

    2023

Pith tools

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