Pith. sign in

REVIEW 3 major objections 6 minor 40 references

Semi-Supervised Semantic Segmentation with High- and Low-level Consistency

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a two-branch semi-supervised model, pairing a GAN-based segmentation branch with a Mean-Teacher classification branch, is the new state of the art in semi-supervised semantic segmentation and gains 5–12 mIoU points…

desk verdict A solid, well-ablated semi-supervised segmentation paper that likely works, but the state-of-the-art claim is under-supported because the only other prior true SSL method (Souly et al.) is never compared quantitatively. read the letter →

arxiv 1908.05724 v1 pith:PTL6D4MO submitted 2019-08-15 cs.CV

Sudhanshu Mittal , Maxim Tatarchenko , Thomas Brox This is my paper · ORCID
classification cs.CV MSC 68T0768T45
keywords semi-supervisedlearningsemanticsegmentationgenerativeadversarialnetworksself-trainingmeanteachermulti-labelclassificationfeaturematchinglow-dataregime
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 aims to show that semantic segmentation can be learned from very few pixel-labeled images when a larger pool of unlabeled images is available. It proposes a two-branch model: a GAN-based branch that sharpens boundaries and coherent surfaces, and a semi-supervised multi-label classification branch that suppresses classes absent from the image. On PASCAL VOC, PASCAL-Context, and Cityscapes, the combined model outperforms the supervised baseline and the previous adversarial semi-supervised method, with the largest gains in the low-label regime. The result would matter because dense pixel annotation is expensive, and a reliable method that exploits unlabeled images could reduce that cost.

What carries the argument

The machinery is a two-branch architecture with late fusion. The s4GAN branch consists of a DeepLabv2 segmentation network acting as generator and a convolutional image-wise discriminator; its losses are cross-entropy on labeled pixels, feature matching on unlabeled predictions, and a self-training loss that converts discriminator-confident predictions into pseudo ground truth. The MLMT branch is a Mean Teacher classifier whose student and teacher networks see differently perturbed images, trained by classification loss plus a consistency loss; at inference a threshold $\tau=0.2$ on the teacher's soft class scores switches off segmentation channels. The key identity the argument relies on is that the discriminator score is a usable confidence measure for pseudo-labeling, which lets unlabeled data enter supervised training without manual labels.

What would settle it

Take a trained s4GAN model and extract the predictions that pass the confidence threshold $\gamma$; compare them against ground truth on a held-out subset of the unlabeled pool. If the pseudo-labels have high error on some classes or the mIoU gain disappears when those confident-but-wrong classes are removed, the self-training mechanism is not doing what the paper assumes.

Watch

Extended reading notes

Core claim

The central claim is that low-level and high-level segmentation artifacts fail in different ways and should be fixed by different branches. The low-level branch (s4GAN) trains the segmentation network as a generator against an image-wise discriminator, replacing the standard GAN loss with a feature-matching loss and adding a self-training term that reuses confident discriminator-approved predictions as pseudo-labels. The high-level branch (MLMT) adapts Mean Teacher to multi-label image classification and uses its class-presence scores to zero out segmentation channels for classes the image does not contain. The paper reports that this combination raises mIoU by roughly 5–12 points over the supervised baseline at 1/50 to 1/4 labeled data on PASCAL VOC, and consistently improves over the prior GAN-based method [15], with ablations attributing the gain to all three loss terms.

Load-bearing premise

The load-bearing premise is that the discriminator's confidence score reliably indicates which unlabeled predictions are correct enough to serve as pseudo ground truth; if the discriminator is confidently wrong about certain structures, self-training will reinforce exactly those errors.

Editorial extensions

If this is right

  • With only 1/50 of PASCAL VOC labeled, the combined model reaches 60.4 mIoU without COCO pre-training, 12.1 points above the supervised baseline and 11.2 above the prior GAN method.
  • The s4GAN branch alone provides most of the gain; the MLMT branch adds a complementary 2 points by removing false positives, but brings no gain on Cityscapes because nearly all classes appear in nearly every image.
  • The method also benefits from extra image-level weak labels: adding 9K weak annotations to 1.4K strong ones yields 69.6 mIoU, beating weakly supervised methods that rely on CRF post-processing.
  • All three generator losses matter: feature matching outperforms the standard GAN loss, and adding self-training gives a further gain, with the discriminator output staying closer to 0.5.

Reading between the lines

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

  • Inference: The self-training step would benefit from direct measurement of pseudo-label accuracy; the paper reports end-task mIoU but never measures how often discriminator-approved predictions are actually correct, so the mechanism supporting the gain remains partly inferred.
  • Inference: Because the MLMT branch is only useful when class presence varies across images, a class-conditional or class-balanced variant could extend the same idea to datasets like Cityscapes where the branch is currently inactive.
  • Inference: The reported stability of feature matching over the standard GAN loss suggests the method could transfer to domains such as medical imaging, where ImageNet pre-training is less helpful and the MLMT branch already shows a larger ROC improvement from scratch.
  • Inference: The approach does not require generated images, only predictions from one network, so it could be combined with consistency regularization at the pixel level (e.g., perturbation-based losses) instead of only at the image level.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes a dual-branch method for semi-supervised semantic segmentation. The first branch (s4GAN) trains a DeepLabv2 segmentation network as a generator in a GAN, using a feature-matching loss and a self-training loss that reuses high-confidence discriminator outputs as pseudo-labels. The second branch (MLMT) is a Mean Teacher multi-label classifier that suppresses false-positive class channels at evaluation time. Experiments on PASCAL VOC 2012, PASCAL-Context, and Cityscapes report consistent mIoU improvements over a supervised DeepLabv2 baseline and over Hung et al., especially at low labeled-data fractions, along with ablations of the loss terms and branch contributions.

Significance. If the empirical results hold, the paper offers a practically useful recipe for semantic segmentation with very few pixel labels, combining adversarial training, feature matching, self-training, and a semi-supervised image-level classifier. The ablations in Tables 5 and 6 support the individual design choices, and the release of source code is a plus for reproducibility. The multi-dataset evaluation and the honest reporting of failure cases are also strengths. The significance is tempered, however, by the absence of a quantitative comparison with Souly et al., one of only two prior true semi-supervised methods named by the paper, and by the lack of variance estimates over random splits.

major comments (3)
  1. [Abstract; Section 2; Tables 1-4] The abstract and Section 4.2.1 claim "new state-of-the-art" in semi-supervised semantic segmentation, but the quantitative comparisons in Tables 1-4 and Figure 2 include only Hung et al. [15] and supervised DeepLabv2 baselines. Section 2 explicitly names Souly et al. [34] as one of only two prior true semi-supervised methods for this task, yet no mIoU value for Souly et al. appears anywhere in the experiments. Because the central claim is competitive superiority, the omission of a direct comparison with a named prior method leaves the claim unsupported. Please add the Souly et al. numbers on the same splits and backbone, or restrict the claim to showing improvement over Hung et al.
  2. [Section 4.1.3; Section 4.2.2; Tables 1-4] All reported mIoU numbers are single runs on a single randomly sampled labeled/unlabeled split, and the hyperparameters λ_fm, λ_st, λ_cons, τ, and γ are selected through cross-validation on the same validation set that is used for the reported final numbers. The threshold experiments in Section 4.2.2 explicitly search over thresholds on the validation set and report the best value. With only one split and no variance estimates, the claimed 5-12 mIoU point gains cannot be distinguished from split luck, and the validation-set tuning may introduce optimism. Please report results over at least three random splits or seeds (mean ± std), and either use a separate held-out set for hyperparameter selection or provide a sensitivity analysis for τ and γ.
  3. [Table 4; Section 4.2.1 (Cityscapes)] On Cityscapes, Table 4 reports only "Ours (s4GAN only)" and no s4GAN+MLMT result, and the text states that the MLMT branch was ineffective for this dataset because most classes appear in almost every image. The abstract nevertheless claims that the dual-branch approach achieves new state-of-the-art on "several standard benchmarks," which implicitly includes Cityscapes. The claim should be qualified, and the full dual-branch model should either be evaluated and reported on Cityscapes or the Cityscapes claim limited to the s4GAN branch. As written, the dual-branch contribution is not demonstrated on that benchmark.
minor comments (6)
  1. [Section 2] In Section 2, the phrase "the two-stage training process of [23]" is a reference error: the two-stage GAN training belongs to Hung et al. [15], not Long et al. [23].
  2. [Equation (3)] In Eq. (3), the condition should be written as D(S(x_u)) ≥ γ, and the definition of the pseudo-label y* (presumably the argmax of S(x_u) per pixel) should be stated explicitly.
  3. [Section 4.2.1] The text says the method gives a "performance increase of 5% to 12%" over the baseline; this is more accurately stated as 5 to 12 mIoU points, to avoid confusion with relative percentage improvements.
  4. [Section 4.1.3] The paper says "Through cross-validation, we find the optimal loss weights" but never describes the procedure: which data splits were used, how many folds, and whether the validation set used for reporting was excluded from that search. Please clarify the protocol.
  5. [Tables 1-5] The label fractions are expressed inconsistently: Table 1 uses 1/50, 1/20, 1/8, 1/4 while Table 5 uses 5:95 and the text sometimes refers to percents. Please standardize the notation.
  6. [Abstract] The abstract's claim of "new state-of-the-art in semi-supervised learning" is broader than the actual task; please qualify it to "semi-supervised semantic segmentation."

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is an empirical benchmark result, not a derivation reduced to its own inputs.

full rationale

The paper's central claim is a set of empirical mIoU numbers on standard benchmarks. The dual-branch model is defined through explicit losses (cross-entropy, feature matching, self-training, and mean-teacher consistency), and its performance is measured against held-out validation data. The self-training loss in Eq. 3 does use pseudo-labels y* generated from the segmentation network's own predictions, but this is self-training by design, a standard semi-supervised learning procedure, not a claim that the model derives a result from itself; the reported improvement is an externally measured quantity on data not used to fit the loss. The paper includes an ablation (Table 6) showing that adding the self-training loss improves mIoU (55.4 to 58.1 at 1/50), providing empirical support for the component. Hyperparameters were tuned on the validation set, which is a possible leakage concern but not a circularity of the derivation. The absence of a quantitative comparison to Souly et al. [34] weakens the 'new state-of-the-art' claim as an evidential matter, but an omitted comparator is not a circular step. There is no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via citation; citations to Salimans et al. and Tarvainen and Valpola reference external, independently established methods. The paper is self-contained against external benchmarks, so no significant circularity is present.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

No new physical entities are introduced. The central claim rests on standard deep learning components, several cross-validated hyperparameters, and assumptions about discriminator confidence and class-presence filtering.

free parameters (5)
  • lambda_fm = 0.1
    Feature matching loss weight, selected by cross-validation (Sec. 4.1.3).
  • lambda_st = 1.0
    Self-training loss weight, selected by cross-validation (Sec. 4.1.3).
  • lambda_cons = 1.0
    Mean Teacher consistency loss weight, selected by cross-validation (Sec. 4.1.3).
  • tau = 0.2
    Class-presence threshold for MLMT fusion, obtained by cross-validation (Sec. 3.3, 4.1.3).
  • gamma = 0.6 for PASCAL datasets, 0.7 for Cityscapes
    Discriminator confidence threshold for self-training, set per dataset (Sec. 4.1.3).
assumptions (3)
  • domain assumption Discriminator score is a valid confidence measure for selecting pseudo-labels.
    Underlies Eq. 3 self-training loss; not independently tested in the paper.
  • domain assumption Mean Teacher consistency regularization transfers from single-label classification to multi-label image classification.
    The MLMT branch (Eq. 6) assumes the student-teacher consistency loss helps multi-label prediction with unlabeled data; only downstream segmentation results are shown.
  • domain assumption A single class-presence threshold applied to all classes removes false positives more often than it removes true positives.
    Eq. 7 zeros entire class channels for a low classifier score; failure cases in Fig. 8 acknowledge true positives can be wrongly removed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Semi-Supervised Semantic Segmentation with High- and Low-level Consistency." pith.science (2026). https://pith.science/paper/PTL6D4MO

@misc{pith2026190805724,
  author       = {Pith},
  title        = {Pith review of: Semi-Supervised Semantic Segmentation with High- and Low-level Consistency},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PTL6D4MO}},
  note         = {Machine review of arXiv:1908.05724}
}
read the original abstract

The ability to understand visual information from limited labeled data is an important aspect of machine learning. While image-level classification has been extensively studied in a semi-supervised setting, dense pixel-level classification with limited data has only drawn attention recently. In this work, we propose an approach for semi-supervised semantic segmentation that learns from limited pixel-wise annotated samples while exploiting additional annotation-free images. It uses two network branches that link semi-supervised classification with semi-supervised segmentation including self-training. The dual-branch approach reduces both the low-level and the high-level artifacts typical when training with few labels. The approach attains significant improvement over existing methods, especially when trained with very few labeled samples. On several standard benchmarks - PASCAL VOC 2012, PASCAL-Context, and Cityscapes - the approach achieves new state-of-the-art in semi-supervised learning.

Figures

Figures reproduced from arXiv: 1908.05724 by the authors.

Figure 1
Figure 1. An image from the PASCAL VOC dataset (a) and its ground [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Semi-supervised Semantic Segmentation: The proposed semi-supervised learning (SSL) approach improves over the baselines even when only little labeled data is available using unlabeled data, shows considerable improvement especially with less than 5% labeled samples. Performance is shown on the PASCAL VOC dataset without (a) and with (b) COCO pre-training. The two branches act in a complementary manner and successful… view at source ↗
Figure 3
Figure 3. Overview of our proposed semi-supervised segmentation approach. The s4GAN branch is a GAN-based model which improves the low-level [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Qualitative results obtained using our semi-supervised segmentation approach on the PASCAL VOC dataset with 5% labeled data without [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results on the PASCAL-Context dataset using 1/8 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results on the Cityscapes dataset using 1/8 labeled [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Ablation study on the PASCAL VOC dataset showing the contribution of the MLMT (d) and the s4GAN (e) branches individually. The s4GAN [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Failure cases. Sometimes, our approach can lead to under [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Qualitative results on the PASCAL-Context dataset using 1/8 [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Qualitative results on the Cityscapes dataset using 1/8 labeled [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 12
Figure 12. Figure 12: ROC curves for CNN-based classification and MT-based semi [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 33 canonical work pages

  1. [15]

    W. Hung, Y. Tsai, Y. Liou, Y. Lin, and M. Yang. Adversarial learning for semi-supervised semantic segmentation. In BMVC, 2018

    work page 2018

  2. [34]

    Souly, C

    N. Souly, C. Spampinato, and M. Shah. Semi supervised semantic segmentation using generative adversarial network. In ICCV, 2017

    work page 2017

  3. [1]

    Ahn and S

    J. Ahn and S. Kwak. Learning pixel-level semantic affinity with image-level supervision for weakly supervised semantic segmen- tation. In CVPR, 2018

    work page 2018

  4. [2]

    Athiwaratkun, M

    B. Athiwaratkun, M. Finzi, P . Izmailov, and A. G. Wilson. Im- proving consistency-based semi-supervised learning with weight averaging. arXiv preprint arXiv:1806.05594 , 2018

    arXiv 2018

  5. [3]

    L. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille. Semantic image segmentation with deep convolutional nets and fully connected crfs. CoRR, abs/1412.7062, 2014

    arXiv 2014

  6. [4]

    L. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille. Deeplab: Semantic image segmentation with deep convo- lutional nets, atrous convolution, and fully connected crfs. TP AMI, 40(4):834–848, 2018

    work page 2018

  7. [5]

    L. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. In ECCV, 2018

    work page 2018

  8. [6]

    Cordts, M

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Be- nenson, U. Franke, S. Roth, and B. Schiele. The cityscapes dataset for semantic urban scene understanding. In CVPR, 2016

    work page 2016

Show all 40 references
  1. [7]

    J. Dai, K. He, and J. Sun. Boxsup: Exploiting bounding boxes to supervise convolutional networks for semantic segmentation. In ICCV, 2015

    2015

  2. [8]

    J. Deng, W. Dong, R. Socher, L. jia Li, K. Li, and L. Fei-fei. Imagenet: A large-scale hierarchical image database. In CVPR, 2009

    2009

  3. [9]

    Everingham, L

    M. Everingham, L. Gool, C. K. Williams, J. Winn, and A. Zis- serman. The pascal visual object classes (voc) challenge. IJCV, 88(2):303–338, 2010

    2010

  4. [10]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio. Generative adversarial nets. In NIPS. 2014. 11

    2014

  5. [11]

    Hariharan, P

    B. Hariharan, P . Arbelaez, L. Bourdev, S. Maji, and J. Malik. Semantic contours from inverse detectors. In ICCV, 2011

    2011

  6. [12]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In CVPR, 2016

    2016

  7. [13]

    S. Hong, H. Noh, and B. Han. Decoupled deep neural network for semi-supervised semantic segmentation. In NIPS. 2015

    2015

  8. [14]

    J. Hu, L. Shen, and G. Sun. Squeeze-and-excitation networks. In CVPR, 2018

    2018

  9. [16]

    Khoreva, R

    A. Khoreva, R. Benenson, J. Hosang, M. Hein, and B. Schiele. Simple does it: Weakly supervised instance and semantic segmen- tation. In CVPR, 2017

    2017

  10. [17]

    D. P . Kingma and J. Ba. Adam: A method for stochastic optimiza- tion. In ICLR, 2015

    2015

  11. [18]

    Laine and T

    S. Laine and T. Aila. Temporal ensembling for semi-supervised learning. In ICLR, 2017

    2017

  12. [19]

    Lecun, L

    Y. Lecun, L. Bottou, Y. Bengio, and P . Haffner. Gradient-based learning applied to document recognition. In Proceedings of the IEEE, pages 2278–2324, 1998

    1998

  13. [20]

    D. Lin, J. Dai, J. Jia, K. He, and J. Sun. Scribblesup: Scribble- supervised convolutional networks for semantic segmentation. In CVPR, 2016

    2016

  14. [21]

    G. Lin, A. Milan, C. Shen, and I. Reid. Refinenet: Multi-path refinement networks for high-resolution semantic segmentation. In CVPR, 2017

    2017

  15. [22]

    T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P . Perona, D. Ramanan, P . Dollr, and C. L. Zitnick. Microsoft COCO: Common objects in context. In ECCV, 2014

    2014

  16. [23]

    J. Long, E. Shelhamer, and T. Darrell. Fully convolutional networks for semantic segmentation. In CVPR, 2015

    2015

  17. [24]

    P . Luc, C. Couprie, S. Chintala, and J. Verbeek. Semantic segmen- tation using adversarial networks. In NIPS Workshops. 2016

    2016

  18. [25]

    A. L. Maas, A. Y. Hannun, and A. Y. Ng. Rectifier nonlinearities improve neural network acoustic models. In ICML Workshop on Deep Learning for Audio, Speech and Language Processing , 2013

    2013

  19. [26]

    Miyato, S

    T. Miyato, S. ichi Maeda, M. Koyama, and S. Ishii. Virtual adversarial training: a regularization method for supervised and semi-supervised learning. TP AMI, 2018

    2018

  20. [27]

    Mottaghi, X

    R. Mottaghi, X. Chen, X. Liu, N.-G. Cho, S.-W. Lee, S. Fidler, R. Urtasun, and A. Yuille. The role of context for object detection and semantic segmentation in the wild. In CVPR, 2014

    2014

  21. [28]

    Oliver, A

    A. Oliver, A. Odena, C. Raffel, E. D. Cubuk, and I. J. Goodfellow. Realistic evaluation of semi-supervised learning algorithms. In ICLR Workshop. 2018

    2018

  22. [29]

    Papandreou, L.-C

    G. Papandreou, L.-C. Chen, K. P . Murphy, and A. L. Yuille. Weakly- and semi-supervised learning of a deep convolutional network for semantic image segmentation. In ICCV, 2015

    2015

  23. [30]

    Paszke, S

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer. Automatic differ- entiation in pytorch. 2017

    2017

  24. [31]

    C. Peng, X. Zhang, G. Yu, G. Luo, and J. Sun. Large kernel matters – improve semantic segmentation by global convolutional network. In CVPR, 2017

    2017

  25. [32]

    P . O. Pinheiro and R. Collobert. From image-level to pixel-level labeling with convolutional networks. In CVPR, 2015

    2015

  26. [33]

    Salimans, I

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, X. Chen, and X. Chen. Improved techniques for training gans. In NIPS. 2016

    2016

  27. [35]

    Srivastava, G

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov. Dropout: A simple way to prevent neural networks from overfitting. JMLR, 15(1):1929–1958, 2014

    1929

  28. [36]

    M. Tang, F. Perazzi, A. Djelouah, I. B. Ayed, C. Schroers, and Y. Boykov. On regularized losses for weakly-supervised cnn segmentation. In ECCV, 2018

    2018

  29. [37]

    Tarvainen and H

    A. Tarvainen and H. Valpola. Mean teachers are better role mod- els: Weight-averaged consistency targets improve semi-supervised deep learning results. In NIPS. 2017

    2017

  30. [38]

    Y. Wei, J. Feng, X. Liang, M.-M. Cheng, Y. Zhao, and S. Yan. Object region mining with adversarial erasing: A simple classification to semantic segmentation approach. In CVPR, 2017

    2017

  31. [39]

    Y. Wei, H. Xiao, H. Shi, Z. Jie, J. Feng, and T. S. Huang. Revisiting dilated convolution: A simple approach for weakly- and semi- supervised semantic segmentation. In CVPR, 2018

    2018

  32. [40]

    Zhang, K

    H. Zhang, K. Dana, J. Shi, Z. Zhang, X. Wang, A. Tyagi, and A. Agrawal. Context encoding for semantic segmentation. In CVPR, 2018

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.