Pith. sign in

REVIEW 4 major objections 4 minor 44 references

Webly Supervised Multi-Label Recognition: Evaluation Benchmark and Dual-Branch Multi-Label Contrastive Learning

T0 review · 4 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Webly supervised multi-label recognition is feasible if noisy keyword labels are corrected at the category level, and the paper's dual-branch contrastive method shows a 6.8-point mAP gain on an 80-category benchmark.

desk verdict Useful benchmark and plausible method, but the headline superiority claim isn't clean because DBMLCL trains with weight decay 0 while baselines use their own, and the label-correction loop lacks an external check. read the letter →

arxiv 2607.20874 v1 pith:E7LG2L7Q submitted 2026-07-23 cs.CV

Zhihua Xu , Zhijing Yang , Yufeng Yang , Tianshui Chen This is my paper · ORCID
classification cs.CV
keywords weblysupervisedlearningmulti-labelrecognitionlabelnoisecontrastiveprototypecorrectionevaluationbenchmarkimageclassification
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 is trying to establish that multi-label image recognition models can be trained on noisy web-search keyword labels instead of clean manual annotations, and that the noise does not have to be fatal. To do this it assembles two large web-retrieved datasets over the 80 and 20 categories of two standard multi-label benchmarks, and measures that the keyword-derived labels are only about 46% precise and 65% recall. The central claim is that a dual-branch network that learns category-specific instance features and category prototypes, then uses the similarity between them to overwrite suspicious labels, outperforms representative existing methods on both datasets: on the 80-category set it raises mAP from 64.6 to 71.4 at 100% data. If this holds, webly supervised multi-label recognition becomes a workable alternative for scaling recognition systems without manual label cost.

What carries the argument

DBMLCL (Dual-Branch Multi-Label Contrastive Learning) is the load-bearing machinery: two branches with unshared parameters, each trained on a disjoint half of the data. Two contrastive losses do the work: an instance contrastive loss makes the two branches' category-level feature vectors agree for positive categories and disagree with a queue of negatives, and a prototype contrastive loss makes each feature vector cling to its own category prototype and repel others. The resulting feature–prototype similarity, averaged over the two branches, is what the label-correction rule uses to retrieve missing labels and delete false positives, with adaptive per-category thresholds updated by momentum.

What would settle it

Take a random sample of web images that were never used for training, have humans label them, and compare the original keyword labels and the model's corrected labels against those human labels. If the corrected labels are not more precise, or not more balanced in precision and recall, than the original keyword labels on this sample, the correction mechanism is not doing what the paper claims. An even more direct test is to train the same network with the correction module disabled and see whether the reported mAP gap disappears.

Watch

Extended reading notes

Core claim

The paper's central claim is that label noise in webly supervised multi-label recognition should be modeled and corrected at the category level rather than at the level of whole images. Each training image is passed through two branches of the same network; each branch produces a feature vector per category, and the network maintains a running prototype vector per category. After a five-epoch warm-up, the model uses its predicted probabilities to flag label entries that look wrong, and then rewrites those entries according to the cosine similarity between the image's category feature and the category prototype. The authors report that this correction loop yields the best mAP, overall F1, and

Load-bearing premise

The load-bearing premise is that, in the label-correction step, after five warm-up epochs the model's predictions and feature–prototype similarities are already reliable enough to overwrite the keyword labels, and that this self-correction makes labels more accurate rather than locking in the model's early mistakes.

Editorial extensions

If this is right

  • Multi-label models can be trained from hundreds of thousands of keyword-tagged web images, with the main remaining cost being compute rather than annotation.
  • Category-level feature–prototype similarity is a practical signal for both recovering missing labels and rejecting false positives in noisy supervision.
  • The benchmark's fixed protocol lets future WS-MLR algorithms be compared on the same training data, same evaluation sets, and same metrics.
  • The reported ablation shows the gain comes from contrastive learning plus correction, not from merely doubling the network branches.
  • A vision-language baseline can achieve high ranking mAP but lower F1 on this data, implying that ranking quality is not enough for usable multi-label predictions.

Reading between the lines

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

  • Because correction is conservative, the method likely under-recovers rare, small, or heavily occluded categories; a testable tweak is to set category-specific recall targets and adjust the similarity thresholds until the precision/recall trade-off matches class frequency.
  • The correction loop has no external clean labels, so threshold drift is a real risk; one cheap safeguard would be to reserve a small human-labeled sample purely to monitor whether flipped labels become more accurate as training proceeds.
  • The two-branch design doubles training cost, but since the 50%/50% split loses little, the same contrastive agreement could be distilled into a single branch at inference, cutting parameters and FLOPs in half.
  • The same category-specific prototype and noise-correction recipe could transfer to other tag-supervised tasks, such as video multi-label recognition or noisy web retrieval, where labels come from queries rather than human annotators.
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 introduces a benchmark for webly supervised multi-label recognition (WS-MLR), comprising two new web-retrieved training sets, Web-COCO and Web-Pascal, aligned with the MS-COCO and Pascal VOC category vocabularies. It re-implements several multi-label recognition baselines under a supposedly common evaluation protocol and proposes a Dual-Branch Multi-Label Contrastive Learning (DBMLCL) framework that combines instance-level and prototype-level contrastive losses with a self-training label-correction module. The label-correction module uses the model's own prediction probabilities and feature-prototype similarities to identify and rewrite noisy keyword-derived labels. Experiments on Web-COCO and Web-Pascal report improved mAP, OF1, and CF1 over the re-implemented baselines at 20%, 40%, and 100% training-data proportions.

Significance. If the benchmark is made available and the results are reproducible, the WS-MLR benchmark itself is a potentially valuable community resource: the datasets are large, category-aligned with widely used test sets, and evaluation on the original manually annotated COCO/VOC validation/test splits is the correct protocol. The paper also quantifies keyword-label noise (precision 46.1%, recall 64.6%), which is useful for future work. However, the central empirical claim is currently weakened by a protocol confound and by a logical gap in the label-correction mechanism. The proposed label correction, as written, cannot remove false-positive labels, and the claimed improvements over baselines are not isolated from a difference in weight decay. These issues are fixable within the manuscript's scope, but the experiments and derivations must be corrected before the claim of 'superior performance' can be considered established.

major comments (4)
  1. [§4.3, Eq. (12), Algorithm 1] The adaptive threshold update in Eq. (12) is restricted to samples with ŷ_i,k = 1. Since θ2 and α2 are initialized to 0 and never receive an update from a sample with ŷ_i,k = 0, θ2 remains 0 throughout training. Because p_i,k is the average of two sigmoid outputs, p_i,k ∈ (0,1), so the condition p_i,k < θ2 in Eq. (8) can never be true. Consequently the algorithm cannot detect or correct any false-positive keyword label, contradicting the claims in §5.4.4 and Figure 9. The update rule must be split into positive and negative cases, or otherwise updated from rectified negatives; otherwise the false-positive-correction component is vacuous. This must be fixed and the experiments rerun because it directly affects the method's central mechanism.
  2. [§5.1.1 and §4.4] The paper claims a unified training strategy for all algorithms, but §4.4 sets weight decay to 0 for DBMLCL while §5.1.1 keeps 'the original settings in each paper' for all baselines. As a result, the Table 1 mAP/OF1/CF1 gaps (e.g., 71.4 vs. 70.3 for CSRA at 100% data) are not attributable to the proposed contrastive learning and label correction alone; they may reflect the regularization difference, which is known to matter strongly when training on noisy labels. Please rerun all Table 1 baselines with weight decay 0, and DBMLCL with the baselines' decays, or otherwise hold this hyperparameter fixed. Report the sensitivity of the main result to this choice.
  3. [§5.2, Tables 1 and 2] All reported results are single-run with no error bars, standard deviations, or significance tests. Webly supervised training is stochastic and the noise level is high; the Table 1 improvements over the second-best method shrink from 3.6 mAP at 20% data to 1.1 mAP at 100% data. Please report at least 3 seeds with means and standard deviations for the main comparisons, or explicitly justify why the differences are robust to training stochasticity.
  4. [§4.3 label-correction loop] The label-correction thresholds are derived from the model's own probability outputs and feature-prototype similarities (Eqs. 8–12) after only a 5-epoch warm-up, with no external clean set used to monitor or halt the self-training loop. If the early model is confidently wrong on a category, corrections will reinforce its errors. The paper does analyze corrected labels on 20,000 manually annotated samples (§5.4.4), which is helpful, but this analysis is post-hoc and not used to select or stop the training. Please add a diagnostic (e.g., correction precision/recall over training epochs on that annotated subset) or compare with fixed-threshold correction and with a variant that uses the clean subset for threshold selection or early stopping, to show the loop does not drift.
minor comments (4)
  1. [Algorithm 1] Line 12 contains a typo: 'if then' should be 'if ... then'.
  2. [Eq. (12)] The notation in Eq. (12) is ambiguous. As written, the update condition set {i | ŷ_i,k = 1} applies to both α and θ, but the text says all thresholds are adaptive. Please clarify the update rule for θ2 and α2.
  3. [Data availability] The abstract and Section 3 give a GitHub URL for datasets/code, while the Data Availability section gives a Baidu Cloud link with a password. Please verify both links and state the license and access terms.
  4. [Figure 8] Please define whether 'retrieved labels' indicates only added positive labels or also removed false positives. Given the threshold issue in Eq. (12), the figure currently cannot support the claim that false-positive labels are corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the empirical claim is grounded on external annotations and the label-correction loop is a standard self-training mechanism rather than a prediction derived from its own input.

full rationale

The paper's headline result is the mAP/OF1/CF1 of DBMLCL on Web-COCO and Web-Pascal, evaluated on the manually annotated MS-COCO validation and Pascal VOC test sets. Those external labels are never used to fit the model's thresholds or corrected labels, so the benchmark comparison is not self-referential. The label-correction module (Eqs. 8-12) does use the model's own probabilities p and prototype similarities s to overwrite web keyword labels, with adaptive thresholds updated from the same signals; this is a self-training loop, not a circular derivation. The corrected labels are not fed back as the evaluation ground truth, and the post-hoc analysis of corrected labels (Figure 8) is checked against 20k manual annotations, making it externally falsifiable. The method adopts SSGRL's feature module via citation [10], but [10] is prior published work and is also re-implemented as a baseline; no load-bearing premise is justified only by a self-citation. The 'unified training strategy' inconsistency over weight decay (§5.1.1: baselines keep original weight decay, §4.4: DBMLCL uses 0) is a legitimate experimental confound but not a circularity: it affects the attribution of the mAP gap, it does not make the result equivalent to its inputs. Consequently no circular step can be exhibited under the required reduction standard.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical or conceptual entities are introduced; the contribution is a dataset and a training procedure. The main free parameters are standard contrastive learning hyperparameters and threshold initializations. The domain assumptions listed are the load-bearing premises for the empirical claims.

free parameters (5)
  • temperature τ = 0.1
    Used in instance and prototype contrastive losses (Eqs. 3,6); set empirically.
  • momentum factor m = 0.999
    For prototype updates and threshold adaptation (Eqs. 5,12).
  • loss balance factors λ1, λ2 = λ1=0.01, λ2=0.05
    Weights for contrastive losses in final objective (Eq. 14).
  • initial thresholds θ1, θ2, α1, α2 = 1, 0, 1, 0
    For noisy label detection and correction (Eqs. 8,11); adaptively updated but initial values are chosen by hand.
  • branch data partition ratio = 50%/50%
    Each branch sees half the training data; chosen for efficiency (Table 5 shows little difference).
assumptions (4)
  • domain assumption Keyword-derived labels from web search are a weak but usable supervision signal for multi-label recognition.
    The entire benchmark and training rely on this; the paper quantifies noise (precision/recall ~46%/65%) but assumes the signal is learnable.
  • domain assumption The model's prediction probabilities and feature-prototype similarities are reliable indicators for detecting and correcting label noise.
    Label correction in §4.3 uses p and s thresholds; if these signals are biased, correction reinforces errors.
  • domain assumption MS-COCO validation and Pascal VOC test are appropriate held-out evaluations for web-trained models.
    Used for all comparisons; no analysis of domain shift or potential web/test image overlap.
  • standard math Binary cross-entropy loss with pseudo-labels is an appropriate training objective.
    Eq. 13; standard choice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Webly Supervised Multi-Label Recognition: Evaluation Benchmark and Dual-Branch Multi-Label Contrastive Learning." pith.science (2026). https://pith.science/paper/E7LG2L7Q

@misc{pith2026260720874,
  author       = {Pith},
  title        = {Pith review of: Webly Supervised Multi-Label Recognition: Evaluation Benchmark and Dual-Branch Multi-Label Contrastive Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E7LG2L7Q}},
  note         = {Machine review of arXiv:2607.20874}
}
read the original abstract

Training deep learning models with freely available web images can reduce their dependence on costly manual annotations. Although webly supervised learning has been widely studied for single-label recognition, its multi-label counterpart remains underexplored, partly due to the lack of unified benchmarks and fair comparison protocols. To address this gap, we construct a benchmark for webly supervised multi-label recognition (WS-MLR), including Web-COCO and Web-Pascal, and re-implement representative baselines under a unified setting. The two datasets cover the same 80 and 20 categories as MS-COCO and Pascal VOC, respectively, and contain about 300 thousand images retrieved from the Internet using category-word combinations as search keywords. We further propose a Dual-Branch Multi-Label Contrastive Learning (DBMLCL) framework, which learns category-specific instance-level and category-level representations together with their similarities to identify and correct noisy labels. Extensive experiments on the benchmark demonstrate that DBMLCL achieves superior performance compared to representative baselines.

Figures

Figures reproduced from arXiv: 2607.20874 by the authors.

Figure 1
Figure 1. Some examples with false-positive and false-negative labels. The keywords are presented on the left and the manually-annotated labels are presented under each image, with false-positive labels in red and false-negative labels in orange. image formats. We then compute a hash value for each valid image file and remove duplicated images with identical hashes. We further select the subset that contains at least one of t… view at source ↗
Figure 2
Figure 2. Precision and recall of pseudo labels in the Web-COCO dataset [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Distribution of image numbers (%) for each category in the Web-COCO dataset [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Precision and recall of pseudo labels in the Web-Pascal dataset. 4. Dual-Branch Multi-Label Contrastive Learning The DBMLCL framework employs two independent branches, each consisting of a backbone followed by a semantic-aware representation learning module. Each branc…
Figure 5
Figure 5. Figure 5: Distribution of image numbers (%) for each category in the Web-Pascal dataset [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Illustration of the DBMLCL framework. It contains two branches, and each of them consists of a feature extractor, a SARL module, and a classifier. The feature extractor extracts features from input images, and then the class-specific features are obtained by the SARL m…
Figure 7
Figure 7. Figure 7: The AP of each category of our proposed framework and the SSGRL baseline [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Precision and recall of the retrieved labels produced by the label correction algorithm [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Some examples of the rectified labels produced by the label correction algorithm, with retrieved labels in blue and discarded labels in red. References [1] Abdelfattah, R., Guo, Q., Li, X., Wang, X., Wang, S., 2023. Cdul: Clip-driven unsupervised learning for multi-lab…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 2 canonical work pages

  1. [1]

    Cdul: Clip-driven unsupervised learning for multi-label image classification, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp

    Abdelfattah, R., Guo, Q., Li, X., Wang, X., Wang, S., 2023. Cdul: Clip-driven unsupervised learning for multi-label image classification, in: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 1348–1357

    2023

  2. [2]

    Laso: Label-set operations networks for multi-label few-shot learning, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp

    Alfassy, A., Karlinsky, L., Aides, A., Shtok, J., Harary, S., Feris, R., Giryes, R., Bronstein, A.M., 2019. Laso: Label-set operations networks for multi-label few-shot learning, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 6548–6557

    2019

  3. [3]

    Multi-label image recognition with two-stream dynamic graph convolution networks

    Cao, P., Chen, P., Niu, Q., 2021. Multi-label image recognition with two-stream dynamic graph convolution networks. Image and Vision Computing 113, 104238. doi:https://doi.org/10.1016/j.imavis.2021.104238. Zhihua Xu et al.:Preprint submitted to ElsevierPage 15 of 17 Webly Supervised Multi-Label Recognition

    arXiv 2021

  4. [4]

    Learning spatial-temporal coherent correlations for speech-preserving facial expression manipulation

    Chen, T., Lin, J., Yang, Z., Qing, C., Wang, G., Lin, L., 2026. Learning spatial-temporal coherent correlations for speech-preserving facial expression manipulation. IEEE Transactions on Pattern Analysis and Machine Intelligence

    2026

  5. [5]

    Knowledge-guided multi-label few-shot learning for general image recognition

    Chen, T., Lin, L., Chen, R., Hui, X., Wu, H., 2022a. Knowledge-guided multi-label few-shot learning for general image recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 1371–1384. doi:10.1109/TPAMI.2020.3025814

    arXiv 2020

  6. [6]

    Heterogeneous semantic transfer for multi-label recognition with partial labels

    Chen, T., Pu, T., Liu, L., Shi, Y., Yang, Z., Lin, L., 2024a. Heterogeneous semantic transfer for multi-label recognition with partial labels. International Journal of Computer Vision

  7. [7]

    Structured semantic transfer for multi-label recognition with partial labels, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp

    Chen, T., Pu, T., Wu, H., Xie, Y., Lin, L., 2022b. Structured semantic transfer for multi-label recognition with partial labels, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp. 339–346. doi:10.1609/aaai.v36i1.19910

    work page doi:10.1609/aaai.v36i1.19910

  8. [8]

    Cross-domain facial expression recognition: A unified evaluation benchmark and adversarial graph learning

    Chen, T., Pu, T., Wu, H., Xie, Y., Liu, L., Lin, L., 2021a. Cross-domain facial expression recognition: A unified evaluation benchmark and adversarial graph learning. IEEE transactions on pattern analysis and machine intelligence

Show all 44 references
  1. [9]

    Dynamiccorrelationlearningandregularizationformulti-labelconfidence calibration

    Chen,T.,Wang,W.,Pu,T.,Qin,J.,Yang,Z.,Liu,J.,Lin,L.,2024b. Dynamiccorrelationlearningandregularizationformulti-labelconfidence calibration. IEEE Transactions on Image Processing

  2. [10]

    Learning semantic-specific graph representation for multi-label image recognition, in: Proceedings of the IEEE International Conference on Computer Vision, pp

    Chen, T., Xu, M., Hui, X., Wu, H., Lin, L., 2019a. Learning semantic-specific graph representation for multi-label image recognition, in: Proceedings of the IEEE International Conference on Computer Vision, pp. 522–531

  3. [11]

    Webly supervised learning of convolutional networks, in: Proceedings of the IEEE international conference on computer vision, pp

    Chen, X., Gupta, A., 2015. Webly supervised learning of convolutional networks, in: Proceedings of the IEEE international conference on computer vision, pp. 1431–1439

    2015

  4. [12]

    Learning graph convolutional networks for multi-label recognition and applications

    Chen, Z., Wei, X.S., Wang, P., Guo, Y., 2021b. Learning graph convolutional networks for multi-label recognition and applications. IEEE Transactions on Pattern Analysis and Machine Intelligence

  5. [13]

    Multi-label image recognition with graph convolutional networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp

    Chen, Z.M., Wei, X.S., Wang, P., Guo, Y., 2019b. Multi-label image recognition with graph convolutional networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 5177–5186

  6. [14]

    Imagenet: A large-scale hierarchical image database, in: Computer Vision and Pattern Recognition, 2009

    Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L., 2009. Imagenet: A large-scale hierarchical image database, in: Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, IEEE. pp. 248–255

    2009

  7. [15]

    A multi-label classification method based on transformer for deepfake detection

    Deng, L., Zhu, Y., Zhao, D., Chen, F., 2024. A multi-label classification method based on transformer for deepfake detection. Image and Vision Computing 152, 105319. doi:https://doi.org/10.1016/j.imavis.2024.105319

    2024

  8. [16]

    Learning a deep convnet for multi-label classification with partial labels, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp

    Durand, T., Mehrasa, N., Mori, G., 2019. Learning a deep convnet for multi-label classification with partial labels, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 647–657

    2019

  9. [17]

    Thepascalvisualobjectclasses(voc)challenge

    Everingham,M.,VanGool,L.,Williams,C.K.,Winn,J.,Zisserman,A.,2010. Thepascalvisualobjectclasses(voc)challenge. International journal of computer vision 88, 303–338

    2010

  10. [18]

    Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

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

    2016

  11. [19]

    Noisy multi-label learning through co-occurrence-aware diffusion

    Hou, S., Ren, Y., Jiang, G., Wang, W., 2025. Noisy multi-label learning through co-occurrence-aware diffusion. Advances in Neural Information Processing Systems. URL:https://mlanthology.org/neurips/2025/hou2025neurips-noisy/. poster presentation

    2025

  12. [20]

    Interactive multi-label cnn learning with partial labels, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Huynh, D., Elhamifar, E., 2020. Interactive multi-label cnn learning with partial labels, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 9423–9432

    2020

  13. [21]

    Classifier-guided clip distillation for unsupervised multi-label classification, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Kim, D., Shim, H., 2025. Classifier-guided clip distillation for unsupervised multi-label classification, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4661–4671

    2025

  14. [22]

    Large loss matters in weakly supervised multi-label classification, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Kim, Y., Kim, J.M., Akata, Z., Lee, J., 2022. Large loss matters in weakly supervised multi-label classification, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14156–14165

    2022

  15. [23]

    Adam: A method for stochastic optimization

    Kingma, D., Ba, J., 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980

    2014 arXiv

  16. [24]

    Mopro: Webly supervised learning with momentum prototypes

    Li, J., Xiong, C., Hoi, S.C., 2020. Mopro: Webly supervised learning with momentum prototypes. arXiv preprint arXiv:2009.07995

    2020 arXiv

  17. [25]

    Webvision database: Visual learning and understanding from web data

    Li, W., Wang, L., Li, W., Agustsson, E., Van Gool, L., 2017. Webvision database: Visual learning and understanding from web data. arXiv preprint arXiv:1708.02862

    2017 arXiv

  18. [26]

    Microsoft coco: Common objects in context, in: European Conference on Computer Vision, Springer

    Lin, T.Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., Zitnick, C.L., 2014. Microsoft coco: Common objects in context, in: European Conference on Computer Vision, Springer. pp. 740–755

    2014

  19. [27]

    Joint multi-label learning and feature extraction for temporal link prediction

    Ma, X., Tan, S., Xie, X., Zhong, X., Deng, J., 2022. Joint multi-label learning and feature extraction for temporal link prediction. Pattern Recognition 121, 108216. doi:https://doi.org/10.1016/j.patcog.2021.108216

    2022

  20. [28]

    Exploringthelimitsofweakly supervised pretraining, in: Proceedings of the European conference on computer vision (ECCV), pp

    Mahajan,D.,Girshick,R.,Ramanathan,V.,He,K.,Paluri,M.,Li,Y.,Bharambe,A.,VanDerMaaten,L.,2018. Exploringthelimitsofweakly supervised pretraining, in: Proceedings of the European conference on computer vision (ECCV), pp. 181–196

    2018

  21. [29]

    Spatial-temporal knowledge-embedded transformer for video scene graph generation

    Pu, T., Chen, T., Wu, H., Lu, Y., Lin, L., 2023. Spatial-temporal knowledge-embedded transformer for video scene graph generation. arXiv preprint arXiv:2309.13237

    2023 arXiv

  22. [30]

    Semanticrepresentationanddependencylearningformulti-labelimagerecognition

    Pu,T.,Yuan,L.,Wu,H.,Chen,T.,Tian,L.,Lin,L.,2022. Semanticrepresentationanddependencylearningformulti-labelimagerecognition. arXiv preprint arXiv:2204.03795

    2022 arXiv

  23. [31]

    Learning transferable visual models from natural language supervision, in: Meila, M., Zhang, T

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I., 2021. Learning transferable visual models from natural language supervision, in: Meila, M., Zhang, T. (Eds.), Proceedings of the 3...

    2021

  24. [32]

    Asymmetric loss for multi-label classification, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp

    Ridnik, T., Ben-Baruch, E., Zamir, N., Noy, A., Friedman, I., Protter, M., Zelnik-Manor, L., 2021. Asymmetric loss for multi-label classification, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 82–91

    2021

  25. [33]

    Deep co-image-label hashing for multi-label image retrieval

    Shen, X., Dong, G., Zheng, Y., Lan, L., Tsang, I.W., Sun, Q.S., 2022. Deep co-image-label hashing for multi-label image retrieval. IEEE Transactions on Multimedia 24, 1116–1126. doi:10.1109/TMM.2021.3119868

    2022

  26. [34]

    Meta-learning for multi-label few-shot classification, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp

    Simon, C., Koniusz, P., Harandi, M., 2022. Meta-learning for multi-label few-shot classification, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp. 3951–3960

    2022

  27. [35]

    Very deep convolutional networks for large-scale image recognition, in: Bengio, Y., Le- Cun, Y

    Simonyan, K., Zisserman, A., 2015. Very deep convolutional networks for large-scale image recognition, in: Bengio, Y., Le- Cun, Y. (Eds.), International Conference on Learning Representations, pp. 1–14. URL:https://mlanthology.org/iclr/2015/ Zhihua Xu et al.:Preprint submitted...

    2015

  28. [36]

    Webly supervised fine-grained recognition: Benchmark datasets and an approach, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp

    Sun, Z., Yao, Y., Wei, X.S., Zhang, Y., Shen, F., Wu, J., Zhang, J., Shen, H.T., 2021. Webly supervised fine-grained recognition: Benchmark datasets and an approach, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 10602–10611

    2021

  29. [37]

    Multi-labelout-of-distributiondetectionviaexploitingsparsityandco-occurrence of labels

    Wang,L.,Huang,S.,Huangfu,L.,Liu,B.,Zhang,X.,2022. Multi-labelout-of-distributiondetectionviaexploitingsparsityandco-occurrence of labels. Image and Vision Computing 126, 104548. doi:https://doi.org/10.1016/j.imavis.2022.104548

    2022

  30. [38]

    Hcp: A flexible cnn framework for multi-label image classification

    Wei, Y., Xia, W., Lin, M., Huang, J., Ni, B., Dong, J., Zhao, Y., Yan, S., 2016. Hcp: A flexible cnn framework for multi-label image classification. IEEE transactions on pattern analysis and machine intelligence 38, 1901–1907

    2016

  31. [39]

    Adahgnn: Adaptive hypergraph neural networks for multi-label image classification, in: Proceedings of the 28th ACM International Conference on Multimedia (ACMMM), pp

    Wu, X., Chen, Q., Li, W., Xiao, Y., Hu, B., 2020. Adahgnn: Adaptive hypergraph neural networks for multi-label image classification, in: Proceedings of the 28th ACM International Conference on Multimedia (ACMMM), pp. 284–293

    2020

  32. [40]

    Multi-labelrecognitioninopendrivingscenariosbasedonbipartite-drivensuperimposeddynamicgraph

    Wu,X.,Chen,S.,2024. Multi-labelrecognitioninopendrivingscenariosbasedonbipartite-drivensuperimposeddynamicgraph. Imageand Vision Computing 149, 105189. doi:https://doi.org/10.1016/j.imavis.2024.105189

    2024

  33. [41]

    Attention-drivendynamicgraphconvolutionalnetworkformulti-labelimagerecognition,in: European Conference on Computer Vision, Springer

    Ye,J.,He,J.,Peng,X.,Wu,W.,Qiao,Y.,2020. Attention-drivendynamicgraphconvolutionalnetworkformulti-labelimagerecognition,in: European Conference on Computer Vision, Springer. pp. 649–665

    2020

  34. [42]

    Deep semantic ranking based hashing for multi-label image retrieval, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    Zhao, F., Huang, Y., Wang, L., Tan, T., 2015. Deep semantic ranking based hashing for multi-label image retrieval, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1556–1564

    2015

  35. [43]

    Towards robust multi-label learning against dirty label noise, in: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, pp

    Zhao, Y., Wang, Y., Wang, Z., Shan, W., Huang, M., Wang, M., Huang, M., Wang, X., 2024. Towards robust multi-label learning against dirty label noise, in: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, pp. 5581–5589. doi:10.24963/ijc...

    2024 doi

  36. [44]

    Residual attention: A simple but effective method for multi-label recognition, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp

    Zhu, K., Wu, J., 2021. Residual attention: A simple but effective method for multi-label recognition, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 184–193. Zhihua Xu et al.:Preprint submitted to ElsevierPage 17 of 17

    2021

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.