Pith. sign in

REVIEW 5 major objections 5 minor 33 references

Benchmarking Federated Learning for Semantic Datasets: Federated Scene Graph Generation

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · 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 to deliver the first benchmark framework that enables federated learning and its evaluation for multi-semantic vision tasks under controlled semantic heterogeneity.

desk verdict Useful first FL benchmark for scene graph generation; the clustering pipeline is clean, but 'controlled semantic heterogeneity' needs quantitative cluster validation and multi-seed results. read the letter →

arxiv 2412.10436 v3 pith:ZISLSVO7 submitted 2024-12-11 cs.CV cs.LG

SeungBum Ha , Taehwan Lee , Jiyoun Lim , Sung Whan Yoon This is my paper · ORCID
classification cs.CVcs.LG
keywords SceneGraphGenerationPanopticFederatedLearningDataHeterogeneitySemanticK-meansClusteringCategoryTensorBenchmark
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

Federated learning benchmarks today are built around single-label classification, so they cannot measure how models cope when each image carries several interlocking semantics. This paper proposes a two-step benchmark process for multi-semantic datasets: first, turn every sample's labels into a category tensor and cluster samples by scene type with K-means; second, distribute those clusters across clients with shard-based or Dirichlet-based partitions to set the level of semantic heterogeneity. As a proof of concept, the authors build a federated Panoptic Scene Graph Generation benchmark and show that four existing PSG methods lose accuracy as the data becomes more semantically non-IID. They also show that momentum-based federated optimizers recover part of that loss, and that the same recipe transfers to the multi-attribute CelebA dataset. The point of the work is to give the field a controllable, privacy-preserving testbed for semantic tasks, not just object labels.

What carries the argument

The central object is the category tensor, a $\mathbb{R}^{13 \times 13 \times 7}$ array for PSG built by mapping each image's subject, object, and predicate into coarse super-classes; each fine-grained label occupies an entry along its axis, so the tensor summarizes the full semantic content of an image in one vector. K-means clustering on the collection of these tensors discovers semantic scene clusters, and the cluster assignment then becomes the single label used by standard FL partitioning schemes. The two partition mechanisms carry the control: shard-based partitioning with parameter $p$ (number of clusters per client) and Dirichlet-based partitioning with concentration $\alpha$ both interpolate smoothly between IID and strongly non-IID semantic distributions.

What would settle it

Run the same shard and Dirichlet partitions with shuffled cluster labels while keeping every client's data quantity identical; if the IID-to-non-IID performance gap persists, the effect is driven by partition sizes or client data volume, not by the semantic content of the clusters. Alternatively, compute cluster purity or silhouette scores on held-out PSG annotations; if the K-means clusters do not separate scene types beyond the 13-object and 7-predicate super-class mapping, the benchmark's claimed semantic heterogeneity is not established.

Watch

Extended reading notes

Core claim

The paper's central claim is that it supplies the first benchmark framework enabling federated learning and its evaluation for multi-semantic vision tasks under controlled semantic heterogeneity. The enabling step is to compress each image's scene graph (objects, subjects, predicates) into a category tensor, run K-means to obtain five semantic clusters (animal scenes, daily human activities, transportation scenes, sports scenes, urban-nature landscapes), equalize cluster sizes to avoid long-tail confounding, and then partition the clusters across clients with either shard assignment or a Dirichlet distribution whose concentration parameter $\alpha$ dials heterogeneity from IID to strongly non-IID. The benchmark's validity rests on the observed performance ordering: centralized learning beats IID federated learning, IID roughly matches random partitioning, and strong non-IID causes clear degradation across IMP, MOTIFS, VCTree, and GPS-Net. On top of that baseline, FedAvgM and FedAdam improve accuracy under semantic heterogeneity, and the same clustering-and-partition recipe on CelebA reproduces the expected slowdown and accuracy loss in non-IID settings.

Load-bearing premise

The load-bearing premise is that the coarse category tensor built from 13 object super-classes and 7 predicate super-classes preserves enough semantic information that K-means clusters correspond to meaningful scene types, so the induced client partitions reflect genuine semantic heterogeneity rather than an artifact of the coarse mapping or the chosen cluster count $n=5$.

Editorial extensions

If this is right

  • Four existing PSG models — IMP, MOTIFS, VCTree, and GPS-Net — all lose Recall and mean-Recall when moving from IID to strongly non-IID semantic partitions, so the benchmark creates a measurable heterogeneity signal rather than a random split.
  • Methods that already handle the long-tailed predicate distribution, notably GPS-Net, are more robust to semantic heterogeneity, suggesting that long-tail robustness and FL heterogeneity robustness are related problems.
  • Server-side momentum (FedAvgM) and adaptive optimization (FedAdam) improve mR@K by about one to three points under shard non-IID, giving FL algorithm developers a semantic task on which to test.
  • Ablation with $n=3$, $n=5$, and $n=10$ clusters preserves the qualitative method ordering and the IID-to-non-IID gap, so the benchmark does not depend on one magical cluster count.
  • On CelebA, the same clustering-plus-partition pipeline produces slower convergence and lower accuracy in non-IID settings, indicating the process is not tied to scene graphs.

Reading between the lines

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

  • Because the category tensor only needs a fixed set of label axes, the same recipe should transfer to other multi-label vision-language tasks such as visual question answering or dense captioning, where one sample carries several semantic claims.
  • A quantitative cluster-quality check — for instance silhouette score on the category tensor or agreement with human scene-type labels — would let future users choose the number of clusters $n$ without relying on qualitative PCA or the small ablation table.
  • If the semantic-heterogeneity signal is caused by the scene-type clusters, then shuffling cluster labels while keeping the same partition sizes should erase the IID-to-non-IID performance gap; this is a direct falsification test implied by the benchmark design.
  • The paper's equalization of cluster sizes removes the natural long-tail imbalance; a natural variant would keep the imbalance and test whether FL aggregation methods that downweight dominant clusters can handle both problems at once.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes a benchmark process for federated learning on multi-semantic vision datasets, with controllable semantic heterogeneity. The process has two steps: (i) discovering semantic clusters by K-means on a category tensor built from multi-label annotations, and (ii) partitioning data across clients via shard-based or Dirichlet-distribution-based strategies. As a proof of concept, the authors construct a federated Panoptic Scene Graph Generation (PSG) benchmark, evaluate four PSG models (IMP, MOTIFS, VCTree, GPS-Net) under different heterogeneity settings, apply FedAvgM and FedAdam to show robustness gains, and extend the clustering idea to CelebA for a binary classification task. The central claim is that this is the first benchmark framework enabling FL evaluation for multi-semantic vision tasks under controlled semantic heterogeneity.

Significance. If the central claim holds, the paper addresses a genuine gap: existing FL benchmarks rely on single-label classification, while real-world tasks such as scene graph generation carry multiple interdependent semantics. The proposed pipeline—cluster multi-label samples into semantic groups, then partition clients along those groups—is a sensible and potentially reusable recipe, and the paper includes code. The extension to CelebA and the comparison of robust FL algorithms are useful additions. However, the empirical validation is currently too weak to establish the central claim: the cluster semantics are not quantitatively validated, the reported IID-to-non-IID degradation is small and sometimes reversed, and all results are single-seed without variance estimates.

major comments (5)
  1. [Section 3.1, Eq. (1)] The semantic meaningfulness of the K-means clusters is load-bearing for the entire benchmark, yet it is validated only by qualitative PCA (Fig. 3), manual cluster descriptions, and a cluster-count ablation (Table 9). No cluster-purity, stability, or separation metrics are reported. Moreover, the category tensor is described only as allocating labels to orthogonal axes (ℝ^{13×13×7} for PSG); the construction details—how counts are computed, whether the tensor is normalized, and how multiple relations per image are aggregated—are omitted. Euclidean K-means on raw counts may be dominated by relation counts and frequent super-classes rather than scene semantics. Please specify the tensor construction and report quantitative cluster-validity measures (e.g., silhouette score, stability across K-means restarts, or agreement with human-assigned scene categories).
  2. [Table 1, Table 9, Section 4.2] The central control claim is that increasing semantic heterogeneity degrades performance monotonically, but the data contain multiple reversals. In Table 1, MOTIFS mR@20 is 4.06 for Dir(α=10) (IID) and 4.28 for Dir(α=0.2) (non-IID), the opposite of the claimed trend. In Table 9, VCTree at n=3 has mR@20 4.73 (IID) vs. 4.72 (non-IID), no gap, and IMP at n=10 has 1.93 (IID) vs. 2.38 (non-IID), reversed. The text acknowledges only the MOTIFS case as an 'outlier' and dismisses it as 'not a considerable amount' without statistical support. Since the benchmark's validity rests on the heterogeneity knob producing controlled effects, these exceptions must be addressed with multi-seed runs and significance tests, or the claim should be weakened to an average trend with documented exceptions.
  3. [Section 4.1, all tables] All experimental results are single-seed with no error bars or significance tests. The reported IID-to-non-IID gaps are small—averaged mR@20 drops of -0.77% (shard) and -0.64% (Dirichlet) in Section 4.2—so sampling noise from the data partition or model initialization could easily account for the observed differences. Please report means and standard deviations over at least 3–5 seeds for the main comparisons, and ideally over multiple realizations of the shard/Dirichlet partitions, so the reader can judge whether the heterogeneity effect is larger than the noise floor.
  4. [Section 3.2, Table 4, Table 6] The cluster-balancing step discards a large fraction of the data: each cluster is subsampled to the smallest cluster size, reducing the PSG training set from 46K to 11K images. This means the IID-vs-non-IID comparison potentially conflates semantic heterogeneity with data scarcity, since each client holds only about 114 images. The imbalanced ablation in Table 6 shows the issue is material: GPS-Net's non-IID mR@20 (8.10) exceeds its IID value (7.38), reversing the claimed trend, and the paper itself notes this is unexpected. Please analyze the IID-vs-non-IID gap as a function of the per-client dataset size, or otherwise justify that subsampling to a common size does not confound the heterogeneity control.
  5. [Appendix C.4, Section 6] The conclusion claims the benchmark shows 'consistent trends with prior FL studies' for participation rates, but Table 8 does not support this. Increasing the number of participating clients from 5 to 20 improves only MOTIFS; IMP, VCTree, and GPS-Net do not consistently improve, and the text in C.4 acknowledges this deviation. This undermines the generalizability assertion and needs either a revised claim or a PSG-specific explanation grounded in the task's long-tailed predicate distribution and model architectures.
minor comments (5)
  1. [Section 3.1, Eq. (1)] The notation in Eq. (1) is garbled; the clustering operator and the indexing of the dataset are not clearly defined. Please rewrite the equation with explicit notation for the set of category tensors over the dataset.
  2. [Section 4.2] There is a typo in the subsection heading: 'non-ID' should be 'non-IID'.
  3. [Figure 3 and Figure 7] The PCA visualizations are described only as 'a visualization via Principal Component Analysis'; please state explicitly what features are projected and how much variance the first two components capture, since the visual separation is used as evidence for semantic clustering.
  4. [Table 5] The communication-cost column entries such as '64(x 1)' are not self-explanatory; please specify the units (e.g., millions of parameters × rounds) and clarify what the parenthetical factors represent.
  5. [Conclusion] The limitations paragraph mentions the difficulty of choosing the number of clusters and the reliance on semantic annotations; this is helpful, but it should also mention the sensitivity of the IID-vs-non-IID gap to model and K shown in Table 9.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark's heterogeneity is induced from data-driven clusters and validated by external performance measurements, not by construction.

full rationale

The paper's claimed derivation chain is: (i) transform multi-semantic annotations into category tensors (Sec. 3.1), (ii) apply K-means clustering to obtain semantic clusters, (iii) partition these clusters across clients via shards or Dirichlet sampling (Sec. 3.2), and (iv) validate by measuring PSG and CelebA performance under IID vs non-IID splits (Secs. 4-5). None of these steps reduces to its own inputs. The clusters are discovered from the annotation tensor, not fitted to any target accuracy; the reported performance gaps are measured on the PSG test set, which is external to the clustering procedure, so the IID-to-non-IID degradation is an empirical outcome rather than a consequence of how the partitions were defined. The paper's acknowledged limitations in the Conclusion (choosing the number of clusters, requiring semantic annotations) are scope limitations, not circular reductions. The self-citations [14] (Lee & Yoon, FedGF) and [18] (Lim et al., Metaverse) appear only in the Related Work section as examples of FL methods and shard-based heterogeneity strategies; they are not invoked to justify the benchmark's central premise or to forbid alternatives. There is no uniqueness theorem imported from the authors' prior work and no ansatz smuggled in via citation. The strongest supporting concern is that some reported numbers do not monotonically degrade with heterogeneity (e.g., MOTIFS mR@20 is 4.06 at Dir(alpha=10) but 4.28 at Dir(alpha=0.2) in Table 1, and IMP at n=10 is 1.93 IID vs 2.38 non-IID in Table 9); this weakens the empirical support for 'controlled semantic heterogeneity,' but it is a correctness risk, not circularity. Under the stated circularity criteria, the correct finding is no significant circularity.

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

The central claim rests on hand-chosen cluster numbers and coarse semantic groupings. The free parameters are experimental choices that shape the heterogeneity, but they are not fitted to a target result. No new theoretical entities are introduced. The axioms are domain assumptions about the validity of the category tensor and the frozen-detector setup.

free parameters (4)
  • Number of semantic clusters K = 5 (3 and 10 in ablations)
    K in K-means is chosen by hand after manual examination of clusters; no principled criterion is given. Ablation in Table 9 shows results vary with K.
  • Super-class mapping size = 13 object categories, 7 predicate categories
    Mapping from PSG fine-grained labels to coarse super-classes is hand-defined; it discards most of the 133 objects and 374 predicates, and cluster semantics depend on this mapping.
  • Cluster balancing sample size m = 2.2K per cluster, 11K total training images (24.5% of 46K)
    Subsampling to the smallest cluster removes cluster imbalance but discards about 75% of data; the random seed for sampling is unspecified.
  • Dirichlet concentration alphas = 10, 1, 0.2
    Hand-picked to span approximately IID to non-IID; no calibration to an actual distance from IID is provided.
assumptions (4)
  • domain assumption Euclidean K-means on category tensors yields semantically meaningful clusters.
    Section 3.1, Eq. 1; supported only by qualitative PCA visualization in Fig. 3, with no quantitative cluster validity or stability metric.
  • domain assumption The 13x13x7 super-class category tensor preserves the semantic distinctions relevant for scene graph heterogeneity.
    Section 3.3(i); discards 120 object classes and 367 predicates, potentially erasing fine-grained semantic differences.
  • domain assumption Freezing the Faster R-CNN detector and training only the relation head is a valid proxy for PSG performance in federated settings.
    Section 4.1; limits evaluation to predicate classification, so claims about scene graph generation in FL are not fully end-to-end.
  • ad hoc to paper The number of clusters K=5 (or 3/10 in ablation) is appropriate for inducing semantic heterogeneity.
    Section 3.3 and Table 9; chosen for interpretability, not by an objective criterion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Benchmarking Federated Learning for Semantic Datasets: Federated Scene Graph Generation." pith.science (2026). https://pith.science/paper/ZISLSVO7

@misc{pith2026241210436,
  author       = {Pith},
  title        = {Pith review of: Benchmarking Federated Learning for Semantic Datasets: Federated Scene Graph Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZISLSVO7}},
  note         = {Machine review of arXiv:2412.10436}
}
read the original abstract

Federated learning (FL) enables decentralized training while preserving data privacy, yet existing FL benchmarks address relatively simple classification tasks, where each sample is annotated with a one-hot label. However, little attention has been paid to demonstrating an FL benchmark that handles complicated semantics, where each sample encompasses diverse semantic information, such as relations between objects. Because the existing benchmarks are designed to distribute data in a narrow view of a single semantic, managing the complicated semantic heterogeneity across clients when formalizing FL benchmarks is non-trivial. In this paper, we propose a benchmark process to establish an FL benchmark with controllable semantic heterogeneity across clients: two key steps are (i) data clustering with semantics and (ii) data distributing via controllable semantic heterogeneity across clients. As a proof of concept, we construct a federated PSG benchmark, demonstrating the efficacy of the existing PSG methods in an FL setting with controllable semantic heterogeneity of scene graphs. We also present the effectiveness of our benchmark by applying robust federated learning algorithms to data heterogeneity to show increased performance. To our knowledge, this is the first benchmark framework that enables federated learning and its evaluation for multi-semantic vision tasks under the controlled semantic heterogeneity. Our code is available at https://github.com/Seung-B/FL-PSG.

Figures

Figures reproduced from arXiv: 2412.10436 by the authors.

Figure 1
Figure 1. Examples of single-semantic (Left: classification) and multi-semantic (Middle: scene graph generation) tasks, and the overview of the federated learning process (Right). distributing data samples to multiple clients by considering the heterogeneity. The simulation results reveal that the methods tailored to tackle the long-tailed problem in the Panoptic Scene Graph Generation (PSG) task, where some objects and predi… view at source ↗
Figure 2
Figure 2. Category Tensor K-means Clustering Pipeline. (a) Construct a graph using the objects, subjects, and predicates of each image. (b) Map each label into the super-classes of fine-grained labels. (c) Convert categorized relations (subject, object, and predicate) into the category tensor. (d) Transform every input image into a category tensor and perform K-Means Clustering. be similar, and collecting such data on the ser… view at source ↗
Figure 3
Figure 3. Semantic clusters of PSG dataset (a visualization via Principal Component Analysis) (a) (b) Predicate GRU Objects GRU Predicate Features Object Features Subject state Weighted Sum Object state Predicate state Outbound Pred. state Weighted Sum Inbound Pred. state Object state Edge message Node message Predicate GRU Objects GRU Object Features Object Features Object Features Bi￾LSTM Bi￾LSTM Bi￾LSTM Object Context “Dog… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Overview of PSG Methods (a) IMP, (b) MOTIFS, (c) VCTree, (d) GPS-Net. communication costs required to reach the same level of performance. From the convergence behavior in [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Convergence behaviors of the balanced dataset for Shard IID (Left) and Shard non-IID (Right) cases 0 25 50 75 100 Rounds 0 2 4 6 8 mR@100 0 25 50 75 100 Rounds 0 2 4 6 8 mR@100 0 25 50 75 100 Rounds 0 2 4 6 8 mR@100 IMP MOTIFS VCTree GPS-Net [PITH_FULL_IMAGE:figures/f…
Figure 6
Figure 6. Figure 6: Convergence behaviors of the balanced dataset for Dirichlet distribution (Left: 𝛼 = 10, Middle: 𝛼 = 1, Right: 𝛼 = 0.2) accounts for half of the total communication cost, denoting that GPS-Net is the resource-efficient scene graph generation method in FL. GPS-Net shows …
Figure 7
Figure 7. Figure 7: Semantic clusters of CelebA dataset the hyperparameter, we set the local learning rate as 0.01, momentum ratio for FedAvgM as 0.7, and global learning rate as 1𝑒 −4 for FedAdam in [1𝑒 −1 , 1𝑒 −2 , 1𝑒 −3 , 1𝑒 −4]. The total number of rounds is 100, and the participation…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 24 canonical work pages

  1. [1]

    Federated learning based on dynamic regularization

    Acar, D.A.E., Zhao, Y., Navarro, R.M., Mattina, M., Whatmough, P.N., Saligrama, V., 2021. Federated learning based on dynamic regularization. arXiv preprint arXiv:2111.04263

    arXiv 2021

  2. [2]

    Personalized federated learning with gaussian processes

    Achituve,I.,Shamsian,A.,Navon,A.,Chechik,G.,Fetaya,E.,2021. Personalized federated learning with gaussian processes. Advances in Neural Information Processing Systems 34, 8392–8406

    work page 2021

  3. [3]

    arXiv preprint arXiv:1812.01097

    Caldas, S., Duddu, S.M.K., Wu, P., Li, T., Konečn`y, J., McMahan, H.B.,Smith,V.,Talwalkar,A.,2018.Leaf:Abenchmarkforfederated settings. arXiv preprint arXiv:1812.01097

    arXiv 2018

  4. [4]

    Scene graphs: A survey of generations and applications

    Chang, X., Ren, P., Xu, P., Li, Z., Chen, X., Hauptmann, A., 2021. Scene graphs: A survey of generations and applications. arXiv preprint arXiv:2104.01111 2

    arXiv 2021

  5. [5]

    Learning of visual relations: The devil is in the tails, in: Proceedings of the IEEE/CVFInternationalConferenceonComputerVision,pp.15404– 15413

    Desai, A., Wu, T.Y., Tripathi, S., Vasconcelos, N., 2021. Learning of visual relations: The devil is in the tails, in: Proceedings of the IEEE/CVFInternationalConferenceonComputerVision,pp.15404– 15413

    work page 2021

  6. [6]

    Sharpness- awareminimizationforefficientlyimprovinggeneralization

    Foret, P., Kleiner, A., Mobahi, H., Neyshabur, B., 2020. Sharpness- awareminimizationforefficientlyimprovinggeneralization. Interna- tional Conference on Learning Representations (ICLR)

    work page 2020

  7. [7]

    Fast r-cnn, in: Proceedings of the IEEE interna- tional conference on computer vision, pp

    Girshick, R., 2015. Fast r-cnn, in: Proceedings of the IEEE interna- tional conference on computer vision, pp. 1440–1448

    work page 2015

  8. [8]

    Measuringtheeffectsofnon- identical data distribution for federated visual classification

    Hsu,T.M.H.,Qi,H.,Brown,M.,2019. Measuringtheeffectsofnon- identical data distribution for federated visual classification. arXiv preprint arXiv:1909.06335

    arXiv 2019

Show all 33 references
  1. [9]

    Local pseudo-attributes for long-tailed recognition

    Kim, D.J., Ke, T.W., Stella, X.Y., 2023. Local pseudo-attributes for long-tailed recognition. Pattern Recognition Letters 172, 51–57

    2023

  2. [10]

    Adam: A method for stochastic optimization

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

    2014 arXiv

  3. [11]

    Panop- tic segmentation, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp

    Kirillov,A.,He,K.,Girshick,R.,Rother,C.,Dollár,P.,2019. Panop- tic segmentation, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 9404–9413

    2019

  4. [12]

    Visual genome: Connecting language and vision using crowdsourced dense imageannotations

    Krishna, R., Zhu, Y., Groth, O., Johnson, J., Hata, K., Kravitz, J., Chen, S., Kalantidis, Y., Li, L.J., Shamma, D.A., et al., 2017. Visual genome: Connecting language and vision using crowdsourced dense imageannotations. Internationaljournalofcomputervision123,32– 73

    2017

  5. [13]

    Cifar-10 (canadian institute for advanced research)

    Krizhevsky, A., Nair, V., Hinton, G., 2019. Cifar-10 (canadian institute for advanced research)

    2019

  6. [14]

    Rethinking the flat minima searching in federated learning, in: Forty-first International Conference on Ma- chine Learning

    Lee, T., Yoon, S.W., 2024. Rethinking the flat minima searching in federated learning, in: Forty-first International Conference on Ma- chine Learning

    2024

  7. [15]

    Deeprelationalself-attentionnetworks forscenegraphgeneration.PatternRecognitionLetters153,200–206

    Li,P.,Yu,Z.,Zhan,Y.,2022. Deeprelationalself-attentionnetworks forscenegraphgeneration.PatternRecognitionLetters153,200–206

    2022

  8. [16]

    On the con- vergenceoffedavgonnon-iiddata

    Li, X., Huang, K., Yang, W., Wang, S., Zhang, Z., 2019. On the con- vergenceoffedavgonnon-iiddata. arXivpreprintarXiv:1907.02189

    2019 arXiv

  9. [17]

    Uncertainty-aware scene graph generation

    Li, X., Wu, T., Zheng, G., Yu, Y., Li, X., 2023. Uncertainty-aware scene graph generation. Pattern Recognition Letters 167, 30–37

    2023

  10. [18]

    Metavers:Meta-learnedversatile representations for personalized federated learning, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp

    Lim,J.H.,Ha,S.,Yoon,S.W.,2024. Metavers:Meta-learnedversatile representations for personalized federated learning, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 2587–2596

    2024

  11. [19]

    Microsoft coco: Common objects in context, in: Computer Vision–ECCV 2014: 13th European Confer- ence, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, 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: Computer Vision–ECCV 2014: 13th European Confer- ence, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, Springer. pp. 740–755

    2014

  12. [20]

    Gps-net: Graph property sensing network for scene graph generation, in: Proceedings of the IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp

    Lin, X., Ding, C., Zeng, J., Tao, D., 2020. Gps-net: Graph property sensing network for scene graph generation, in: Proceedings of the IEEE/CVFConferenceonComputerVisionandPatternRecognition, pp. 3746–3753

    2020

  13. [21]

    Pattern Recognition Letters 145, 187–193

    Liu,A.A.,Wang,Y.,Xu,N.,Liu,S.,Li,X.,2021.Scene-graph-guided message passing network for dense captioning. Pattern Recognition Letters 145, 187–193

    2021

  14. [22]

    Personalized federated learning on long-tailed data via knowledge distillation and generated features

    Lv, F., Qian, P., Lu, Y., Wang, H., 2024. Personalized federated learning on long-tailed data via knowledge distillation and generated features. Pattern Recognition Letters 186, 178–183

    2024

  15. [23]

    McMahan, B., Moore, E., Ramage, D., Hampson, S., y Arcas, B.A.,

  16. [24]

    Adaptive federated optimization

    Reddi,S.,Charles,Z.,Zaheer,M.,Garrett,Z.,Rush,K.,Konečn `y,J., Kumar, S., McMahan, H.B., 2020. Adaptive federated optimization. arXiv preprint arXiv:2003.00295

    2020 arXiv

  17. [25]

    Federated learning with l1 regularization

    Shi, Y., Zhang, Y., Zhang, P., Xiao, Y., Niu, L., 2023. Federated learning with l1 regularization. Pattern Recognition Letters 172, 15– 21

    2023

  18. [26]

    Learning to compose dynamic tree structures for visual contexts, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition, pp

    Tang, K., Zhang, H., Wu, B., Luo, W., Liu, W., 2019. Learning to compose dynamic tree structures for visual contexts, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition, pp. 6619–6628

    2019

  19. [27]

    5410–5419

    Xu,D.,Zhu,Y.,Choy,C.B.,Fei-Fei,L.,2017.Scenegraphgeneration byiterativemessagepassing,in:ProceedingsoftheIEEEconference on computer vision and pattern recognition, pp. 5410–5419. S. Ha et al.:Accepted for publication in Pattern Recognition Letters. Page 8 of 15 Federated Scene...

    2017

  20. [28]

    Panoptic scene graph generation, in: European Conference on Com- puter Vision, Springer

    Yang, J., Ang, Y.Z., Guo, Z., Zhou, K., Zhang, W., Liu, Z., 2022. Panoptic scene graph generation, in: European Conference on Com- puter Vision, Springer. pp. 178–196

    2022

  21. [29]

    Neural motifs: Scenegraphparsingwithglobalcontext,in:ProceedingsoftheIEEE conference on computer vision and pattern recognition, pp

    Zellers, R.,Yatskar, M., Thomson, S.,Choi, Y., 2018. Neural motifs: Scenegraphparsingwithglobalcontext,in:ProceedingsoftheIEEE conference on computer vision and pattern recognition, pp. 5831– 5840

    2018

  22. [30]

    Panoptic segmentation-based semantic embedding matching model for scene graph generation

    Zhao, M., Zhang, J., 2025. Panoptic segmentation-based semantic embedding matching model for scene graph generation. Pattern Recognition Letters 193, 56–63

    2025

  23. [31]

    sky" or “grass

    Zhu, H., Xu, J., Liu, S., Jin, Y., 2021. Federated learning on non-iid data: A survey. Neurocomputing 465, 371–390. S. Ha et al.:Accepted for publication in Pattern Recognition Letters. Page 9 of 15 Federated Scene Graph Generation Table 4 Dataset Information Data Amount PSG d...

    2021

  24. [33]

    Prior research mentioned that the key point of GPS-Net explicitly models the direction of predicates, which is why it does not perform well on PSG dataset

    is lower than the performance on the balanced dataset. Prior research mentioned that the key point of GPS-Net explicitly models the direction of predicates, which is why it does not perform well on PSG dataset. However, when S. Ha et al.:Accepted for publication in Pattern Rec...

  25. [2017]

    1273–1282

    Communication-efficient learning of deep networks from decentralizeddata,in:Artificialintelligenceandstatistics,PMLR.pp. 1273–1282

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.