Pith. sign in

REVIEW 3 major objections 5 minor 63 references

Dynamic Graph Message Passing Networks

T0 review · 3 major / 5 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 This paper claims that long-range visual context can be captured by a single message-passing layer that dynamically samples a small node neighborhood, predicts input-specific filter weights and affinities, and thereby beats fully…

desk verdict A credible and well-ablated efficiency claim against Non-local, but the adaptive sampling story is weaker than the paper's framing; the 1.3 mIoU dynamic-sampling gain lacks a capacity-matched control. read the letter →

arxiv 1908.06955 v5 pith:SSSSURFO submitted 2019-08-19 cs.CV cs.LG

Li Zhang , Dan Xu , Anurag Arnab , Philip H.S. Torr This is my paper · ORCID
classification cs.CVcs.LG
keywords dynamicgraphmessagepassinglong-rangedependenciesadaptivenodesamplingsemanticsegmentationinstanceobjectdetectionnon-localnetworksrandomwalk
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

Fully connected graph message passing, as in Non-local blocks, models every pairwise relation and costs quadratically in feature count. This paper claims that a single message-passing layer can instead sample a small neighborhood around each node, adapt those samples with input-conditioned random walks, and predict both the filter weights and the affinities used to propagate information, all in one module. The proposed DGMN layer is shown to improve semantic segmentation, object detection, and instance segmentation over strong baselines, and to surpass a fully connected Non-local model on Cityscapes while using about a third of its FLOPs. If the result is right, long-range context for dense prediction does not require all-to-all attention; an input-dependent sparse graph is enough.

What carries the argument

The central mechanism is a dynamic graph message passing (DGMN) layer with two coupled dynamic properties. First, node sampling: uniform sampling at multiple rates gives an initial neighborhood, and a learned random walk (Eq. 4, implemented as a $3\times 3$ convolution) moves each sample to a data-dependent location. Second, message conditioning: the same sampled nodes predict node-specific grouped filter weights and an affinity matrix (Eqs. 6-7) used to weight and transform the sampled features. The message update (Eq. 5) is the object doing the argument's work: it is the single place where dynamic sampling, dynamic filters, and dynamic affinities combine, and the ablation studies attribute gains to each component.

What would settle it

Train the full DGMN and a control in which the predicted offsets are replaced by randomized offsets drawn from the same spatial distribution, or by the mean offset, while everything else is kept fixed. If the control matches the full model on Cityscapes validation, or if the learned offsets are shown to be nearly input-independent, then learned sampling is not the source of the claimed long-range gain.

Watch

Extended reading notes

Core claim

The central claim is that efficiency and accuracy need not be traded off by choosing between locally and fully connected graphs. Interpreting a feature map as graph nodes, DGMN uses a set of uniform sampling rates to generate $S$ small neighborhoods per node, then predicts position-specific random-walk offsets from the sampled features; these offsets relocate the sampled nodes before message passing. On the relocated nodes, a matrix transformation predicts grouped dynamic filter weights and an affinity, normalized along the channel axis, and the message for node $i$ is $\sum_q \sum_{j\in\mathcal N_q(i)} \beta_q A'^q_{i,j}\varrho(h'_j\mid V,j,\Delta d^q_j) w^q_j$. The paper claims this formulation keeps a large receptive field, removes redundant connections, and outperforms fully connected non-local message passing on three tasks while using substantially fewer FLOPs and parameters.

Load-bearing premise

The learned offset that moves each sampled location must actually pick more useful spots than the original grid; if the offsets merely deform the uniform neighborhood, the dynamic-sampling gain collapses and the method reduces to a deformable-convolution variant.

Editorial extensions

If this is right

  • Fully connected non-local graphs can be replaced by smaller sampled graphs: DGMN reaches Cityscapes validation mIoU 80.4 versus 79.0 for Non-local with about 33% of Non-local's FLOPs.
  • The layer is modular: inserting DGMN into a Mask R-CNN backbone improves COCO APbox to 40.2 and APmask to 36.0 over 37.8 and 34.4.
  • Dynamic sampling and dynamic weights/affinities are complementary: ablations show each contributes, with dynamic sampling adding 1.3 mIoU over the already strong DA+DW variant.
  • The same module generalizes across backbones ResNet-50, ResNet-101 and ResNeXt-101 on the COCO test-dev set.

Reading between the lines

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

  • A testable extension is to replace the $3\times 3$ random-walk predictor with a wider or multi-scale context; the reported gains attribute long-range behavior to this predictor, but its local receptive field makes the long-range claim stronger than the architecture alone guarantees.
  • If dynamic sampling works by selecting semantically relevant locations, the same layer should transfer to sparse or irregular graph inputs, where uniform grids are unavailable.
  • The channel-wise softmax affinity and grouped dynamic filters could be interpreted as a structured low-rank attention mechanism, inviting direct comparison with sparse attention using learned indices on sequences or point clouds.
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 / 5 minor

Summary. The paper proposes Dynamic Graph Message Passing Networks (DGMN), a modular layer that replaces fully-connected non-local message passing with a sparse, input-conditioned scheme. For each node, a small set of neighbours is first selected by uniform sampling at multiple rates, then refined by learned random-walk offsets; the sampled nodes are used to predict node-conditioned filter weights and affinities that propagate information into the node update. The authors evaluate the module on Cityscapes semantic segmentation and COCO detection/instance segmentation, reporting that DGMN outperforms Non-local while using about 33% of its FLOPs on Cityscapes, and improves on Mask R-CNN baselines across ResNet-50/101/ResNeXt-101 backbones. The central claim is that input-conditioned sampling of a small neighbourhood, together with dynamically predicted weights and affinities, can capture long-range context more efficiently and accurately than fully-connected attention.

Significance. If the central claim is correct, the paper is a valuable contribution: it shows that a single, plug-in layer can replace quadratic-complexity attention with a sparse, content-adaptive message-passing scheme while improving accuracy, and it validates the approach on three tasks and several backbones. The modular instantiation and the extensive comparisons to Non-local, GCNet, CCNet, and deformable convolution are useful to the community. The paper also includes honest ablations of the DA, DW, and DS components and reports the effect of training/inference strategies in the appendix. However, the specific benefit of the ``dynamic sampling'' component is not cleanly isolated, and the formulation contains a notational/implementation mismatch that needs to be resolved before the contribution can be fully assessed.

major comments (3)
  1. [Table 1 and Sec. 4.2]
  2. [Sec. 3.3, Eq. (4) vs Sec. 3.5]
  3. [All experimental tables (Tables 1-4)]
minor comments (5)
  1. [Appendix, Table 5 caption]
  2. [Sec. 3.3, Eq. (4)]
  3. [Sec. 3.6]
  4. [Eq. (2)]
  5. [Sec. 4.2]

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are external empirical comparisons, and its self-citations are non-load-bearing.

full rationale

The paper's central claims are empirical comparisons on Cityscapes and COCO evaluation servers, so no derivation-from-fitted-parameter circularity applies. DGMN is defined by Eqs. 4-6 (random walk offsets, dynamic filters and affinities) implemented with convolutional layers; the claimed improvements are measured against external baselines such as Non-local and Mask R-CNN, not derived from a fitted quantity that is then renamed as a prediction. The ablation in Table 1 treats sampling rates and group counts as transparent hyperparameters and reports the effect of adding dynamic sampling; this is model selection, not a statistically forced prediction of a fitted value. The only self-citations are references [21] and [31] in Sec. 3.3, used as general motivation that long-range context is important; they do not supply a theorem, forbid alternatives, or carry the load of the accuracy claims, so they do not create circularity. No uniqueness theorem is invoked, and no ansatz is smuggled in via citation. Each load-bearing component is evaluated against held-out public benchmarks with standard protocols. Therefore no step in the claimed derivation chain is equivalent by construction to its input.

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

The central claim rests on four premises: the differentiability of bilinear sampling, the adequacy of a small sampled neighborhood, the sufficiency of a single message passing iteration, and the ability of a local 3x3 offset predictor to produce globally useful sampling. Of these, the last is the most fragile and the least quantified. The main tuned quantities are the sampling rate set and the number of filter groups, both chosen by validation search and both shown to shift the headline numbers.

free parameters (4)
  • Sampling rate set phi = {1,6,12,24,36} on Cityscapes; {1,4,8,12} on COCO.
    Hand-chosen per dataset; Table 6 shows mIoU rises from 79.1 (phi={1}) to 80.4 (full set) on Cityscapes validation, so this choice directly shapes the central efficiency-accuracy claim.
  • Number of filter groups G = 4
    Hand-chosen; Table 7 shows APbox from 39.5 (G=2) to 40.2 (G=4), so the reported COCO gains partly depend on this setting.
  • Message passing iterations T = 1
    Fixed for efficiency following prior work; the paper does not ablate it, so all gains are reported for a single-iteration module.
  • Neighborhood size K = 9 (3x3)
    Structural choice that scales FLOPs per node; no experiment varies K, so sensitivity to this constant is unknown.
assumptions (4)
  • standard math Bilinear sampling rho (Eq. 5, Sec 3.3) is differentiable and lets gradients flow to the predicted walks so the sampling locations can be learned.
    Assumes the feature map is locally continuous so interpolating features at fractional coordinates is a meaningful training signal; the same assumption underlies deformable convolution [13, 63] and Spatial Transformer Networks [23].
  • domain assumption A set of K sampled nodes per receiving node is enough to approximate the full-graph message distribution (Sec 3.3, Eq. 1).
    This is the core approximation of the method; if K nodes miss crucial context, the accuracy relative to a fully-connected graph would fall, but the paper only tests specific K=9 configurations.
  • domain assumption T=1 message passing iteration is sufficient for feature refinement (Sec 3.5).
    Adopted from prior work [51, 8, 33, 59] for efficiency; no ablation over T is reported, so the results are conditional on this choice.
  • domain assumption Walks predicted from a 3x3 convolution over the uniform-sampled neighborhood yield globally relevant sampling locations (Eq. 4, Sec 3.5).
    The offset predictor is local, yet the paper describes it as adaptive long-range sampling; this is the weakest link and is supported only by qualitative visualizations and a single ablation row.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dynamic Graph Message Passing Networks." pith.science (2026). https://pith.science/paper/SSSSURFO

@misc{pith2026190806955,
  author       = {Pith},
  title        = {Pith review of: Dynamic Graph Message Passing Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SSSSURFO}},
  note         = {Machine review of arXiv:1908.06955}
}
read the original abstract

Modelling long-range dependencies is critical for scene understanding tasks in computer vision. Although CNNs have excelled in many vision tasks, they are still limited in capturing long-range structured relationships as they typically consist of layers of local kernels. A fully-connected graph is beneficial for such modelling, however, its computational overhead is prohibitive. We propose a dynamic graph message passing network, that significantly reduces the computational complexity compared to related works modelling a fully-connected graph. This is achieved by adaptively sampling nodes in the graph, conditioned on the input, for message passing. Based on the sampled nodes, we dynamically predict node-dependent filter weights and the affinity matrix for propagating information between them. Using this model, we show significant improvements with respect to strong, state-of-the-art baselines on three different tasks and backbone architectures. Our approach also outperforms fully-connected graphs while using substantially fewer floating-point operations and parameters. The project website is http://www.robots.ox.ac.uk/~lz/dgmn/

Figures

Figures reproduced from arXiv: 1908.06955 by the authors.

Figure 1
Figure 1. Contextual information is crucial for complex scene understanding tasks. To recognise the “boathouse”, one needs to consider the “boat” and the “water” next to it. Fully-connected message passing models (a) are able to obtain this information, but are prohibitively expensive. Furthermore, they capture a lot of redundant information (i.e.“trees” and “sky”). Locally-connected models (b) are more efficient, but miss ou… view at source ↗
Figure 2
Figure 2. Overview of our proposed dynamic graph message passing network (DGMN). The neighbourhood used to update the feature representation of each node (we show a single node with a red square) is predicted dynamically conditioned on each input. This is done by first uniformly sampling (denoted by “US”) a set of S neighbourhoods around each node. Each neighbourhood contains K (e.g. 3 × 3) sampled nodes. Here, the blue nodes… view at source ↗
Figure 3
Figure 3. Schematic illustration of the proposed dynamic message passing calculation (DMC) module. The small red square indicates the receiving node whose message is calculated from its neighbourhood, i.e. the sampled K (e.g. 3 × 3) features nodes. The module accepts a feature map as input and produces its corresponding message map. The symbol ∗ denotes group convolution operation using the dynamically predicted and position … view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visualisation of the nodes sampled via learning the random walks with our network (trained for instance segmentation on COCO). The red point indicates a receiving node i. Different colour families (i.e. yellow and blue) indicate the learned position specific weights an…
Figure 5
Figure 5. Figure 5: Validation curves of APbox and APmask on COCO for Mask-RCNN baseline, Non-local and the proposed DGMN. The number of training epochs is 90K. in FLOPs and parameters, showing the benefit of using predicted dynamic affinities for reweighting the messages in message passi…
Figure 6
Figure 6. Figure 6: Qualitative examples of our results for semantic segmentation on Cityscapes (first row), and object detection and instance segmentation on COCO (second row) Backbone mIoU (%) PSPNet [60] ResNet 101 78.4 PSANet [61] ResNet 101 80.1 DenseASPP [55] DenseNet 161 80.6 GloRe…
Figure 7
Figure 7. Figure 7: Qualitative results of the Dilated FCN baseline [ [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Qualitative examples of the instance segmentation task on the COCO validation dataset. The odd rows are the results [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: More qualitative examples of the instance segmentation task on the COCO validation dataset. The odd rows are the [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 59 canonical work pages

  1. [1]

    Higher order conditional random fields in deep neural networks

    Anurag Arnab, Sadeep Jayasumana, Shuai Zheng, and Philip HS Torr. Higher order conditional random fields in deep neural networks. In ECCV, 2016. 2

    work page 2016

  2. [2]

    Conditional random fields meet deep neural networks for semantic segmentation: Combining probabilistic graphical models with deep learning for structured prediction

    Anurag Arnab, Shuai Zheng, Sadeep Jayasumana, Bernardino Romera-Paredes, M˚ans Larsson, Alexander Kirillov, Bogdan Savchynskyy, Carsten Rother, Fredrik Kahl, and Philip HS Torr. Conditional random fields meet deep neural networks for semantic segmentation: Combining probabilistic graphical models with deep learning for structured prediction. IEEE Signal Pr...

    work page 2018

  3. [3]

    Geodesic matting: A frame- work for fast interactive image and video segmentation and matting

    Xue Bai and Guillermo Sapiro. Geodesic matting: A frame- work for fast interactive image and video segmentation and matting. IJCV, 2009. 1

    work page 2009

  4. [4]

    Gcnet: Non-local networks meet squeeze-excitation networks and beyond

    Yue Cao, Jiarui Xu, Stephen Lin, Fangyun Wei, and Han Hu. Gcnet: Non-local networks meet squeeze-excitation networks and beyond. arXiv, 2019. 7

    work page 2019

  5. [5]

    Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L. Yuille. Semantic image segmen- tation with deep convolutional nets and fully connected CRFs. In ICLR, 2015. 2, 9

    work page 2015

  6. [6]

    Rethinking atrous convolution for semantic image segmentation

    Liang-Chieh Chen, George Papandreou, Florian Schroff, and Hartwig Adam. Rethinking atrous convolution for semantic image segmentation. arXiv, 2017. 6, 7, 8, 9

    work page 2017

  7. [7]

    Encoder-decoder with atrous separable convolution for semantic image segmentation

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. In ECCV, 2018. 9

    work page 2018

  8. [8]

    Graph-based global reasoning networks

    Yunpeng Chen, Marcus Rohrbach, Zhicheng Yan, Shuicheng Yan, Jiashi Feng, and Yannis Kalantidis. Graph-based global reasoning networks. In CVPR, 2019. 5, 8, 9

    work page 2019

Show all 63 references
  1. [9]

    Xception: Deep learning with depthwise separable convolutions

    Franc ¸ois Chollet. Xception: Deep learning with depthwise separable convolutions. In CVPR, 2017. 5, 9

    2017

  2. [10]

    The cityscapes dataset for semantic urban scene understanding

    Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. The cityscapes dataset for semantic urban scene understanding. In CVPR, 2016. 1, 2, 6, 9

    2016

  3. [11]

    Boxsup: Exploit- ing bounding boxes to supervise convolutional networks for semantic segmentation

    Jifeng Dai, Kaiming He, and Jian Sun. Boxsup: Exploit- ing bounding boxes to supervise convolutional networks for semantic segmentation. In ICCV, 2015. 9

    2015

  4. [12]

    R-fcn: Object detection via region-based fully convolutional networks

    Jifeng Dai, Yi Li, Kaiming He, and Jian Sun. R-fcn: Object detection via region-based fully convolutional networks. In NeurIPS, 2016. 11

    2016

  5. [13]

    Deformable convolutional networks

    Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. Deformable convolutional networks. In ICCV, 2017. 2, 5, 6

    2017

  6. [14]

    Dssd: Deconvolutional single shot detector

    Cheng-Yang Fu, Wei Liu, Ananth Ranga, Ambrish Tyagi, and Alexander C Berg. Dssd: Deconvolutional single shot detector. arXiv, 2017. 11

    2017

  7. [15]

    Dual attention network for scene segmentation

    Jun Fu, Jing Liu, Haijie Tian, Zhiwei Fang, and Hanqing Lu. Dual attention network for scene segmentation. In CVPR,

  8. [16]

    Neural message passing for quantum chemistry

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. Neural message passing for quantum chemistry. In ICML, 2017. 1, 3

    2017

  9. [17]

    Accurate, large minibatch sgd: Training imagenet in 1 hour

    Priya Goyal, Piotr Doll ´ar, Ross Girshick, Pieter Noord- huis, Lukasz Wesolowski, Aapo Kyrola, Andrew Tulloch, Yangqing Jia, and Kaiming He. Accurate, large minibatch sgd: Training imagenet in 1 hour. In arXiv, 2017. 9

    2017

  10. [18]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. In NeurIPS, 2017. 1, 2

    2017

  11. [19]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Gir- shick. Mask r-cnn. In ICCV, 2017. 6, 10, 11, 12, 13

    2017

  12. [20]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, 2016. 1, 6, 11

    2016

  13. [21]

    Strip pooling: Rethinking spatial pooling for scene parsing

    Qinbin Hou, Li Zhang, Ming-Ming Cheng, and Jiashi Feng. Strip pooling: Rethinking spatial pooling for scene parsing. In CVPR, 2020. 3

    2020

  14. [22]

    Ccnet: Criss-cross attention for semantic segmentation

    Zilong Huang, Xinggang Wang, Lichao Huang, Chang Huang, Yunchao Wei, and Wenyu Liu. Ccnet: Criss-cross attention for semantic segmentation. In ICCV, 2019. 7, 8

    2019

  15. [23]

    Spatial transformer networks

    Max Jaderberg, Karen Simonyan, Andrew Zisserman, and Ko- ray Kavukcuoglu. Spatial transformer networks. In NeurIPS,

  16. [24]

    Dynamic filter networks

    Xu Jia, Bert De Brabandere, Tinne Tuytelaars, and Luc V Gool. Dynamic filter networks. In NeurIPS, 2016. 2

    2016

  17. [25]

    Semi-supervised classifi- cation with graph convolutional networks

    Thomas N Kipf and Max Welling. Semi-supervised classifi- cation with graph convolutional networks. In ICLR, 2017. 1, 3

    2017

  18. [26]

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton. Im- ageNet classification with deep convolutional neural networks. In NeurIPS, 2012. 1

    2012

  19. [27]

    Efficient inference in fully connected crfs with gaussian edge potentials

    Philipp Krhenbhl and Vladlen Koltun. Efficient inference in fully connected crfs with gaussian edge potentials. NeurIPS,

  20. [28]

    Cornernet: Detecting objects as paired keypoints

    Hei Law and Jia Deng. Cornernet: Detecting objects as paired keypoints. In ECCV, 2018. 10, 11

    2018

  21. [29]

    Sampling from large graphs

    Jure Leskovec and Christos Faloutsos. Sampling from large graphs. In SIGKDD, 2006. 3

    2006

  22. [30]

    Holistic, instance-level human parsing

    Qizhu Li, Anurag Arnab, and Philip HS Torr. Holistic, instance-level human parsing. In BMVC, 2017. 9

    2017

  23. [31]

    Global aggregation then local dis- tribution in fully convolutional networks

    Xiangtai Li, Li Zhang, Ansheng You, Maoke Yang, Kuiyuan Yang, and Yunhai Tong. Global aggregation then local dis- tribution in fully convolutional networks. In BMVC, 2019. 3

    2019

  24. [32]

    Gated graph sequence neural networks

    Yujia Li, Daniel Tarlow, Marc Brockschmidt, and Richard Zemel. Gated graph sequence neural networks. In ICLR,

  25. [33]

    Deeply learning the messages in message passing inference

    Guosheng Lin, Chunhua Shen, Ian Reid, and Anton van den Hengel. Deeply learning the messages in message passing inference. In NeurIPS, 2015. 5

    2015

  26. [34]

    Girshick, Kaiming He, Bharath Hariharan, and Serge J

    Tsung-Yi Lin, Piotr Dollr, Ross B. Girshick, Kaiming He, Bharath Hariharan, and Serge J. Belongie. Feature pyramid networks for object detection. In CVPR, 2017. 11

    2017

  27. [35]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Doll´ar. Focal loss for dense object detection. In ICCV,

  28. [36]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, 2014. 2, 6, 9

    2014

  29. [37]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, 2014. 9

    2014

  30. [38]

    Ssd: Single shot multibox detector

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C Berg. Ssd: Single shot multibox detector. In ECCV, 2016. 11

    2016

  31. [39]

    maskrcnn-benchmark: Fast, modular reference implementation of Instance Seg- mentation and Object Detection algorithms in PyTorch

    Francisco Massa and Ross Girshick. maskrcnn-benchmark: Fast, modular reference implementation of Instance Seg- mentation and Object Detection algorithms in PyTorch. https://github.com/facebookresearch/ maskrcnn-benchmark, 2018. 6, 9, 12, 13

    2018

  32. [40]

    The role of context in object recognition

    Aude Oliva and Antonio Torralba. The role of context in object recognition. Trends in cognitive sciences, 2007. 1, 3

    2007

  33. [41]

    Libra r-cnn: Towards balanced learning for object detection

    Jiangmiao Pang, Kai Chen, Jianping Shi, Huajun Feng, Wanli Ouyang, and Dahua Lin. Libra r-cnn: Towards balanced learning for object detection. In CVPR, 2019. 10, 11

    2019

  34. [42]

    Full-resolution residual networks for semantic segmentation in street scenes

    Tobias Pohlen, Alexander Hermans, Markus Mathias, and Bastian Leibe. Full-resolution residual networks for semantic segmentation in street scenes. In CVPR, 2017. 9

    2017

  35. [43]

    Objects in context

    Andrew Rabinovich, Andrea Vedaldi, Carolina Galleguillos, Eric Wiewiora, and Serge Belongie. Objects in context. In ICCV, 2007. 1, 3

    2007

  36. [44]

    Yolov3: An incremental improvement

    Joseph Redmon and Ali Farhadi. Yolov3: An incremental improvement. arXiv, 2018. 11

    2018

  37. [45]

    In-place activated batchnorm for memory-optimized training of dnns

    Samuel Rota Bul `o, Lorenzo Porzi, and Peter Kontschieder. In-place activated batchnorm for memory-optimized training of dnns. In CVPR, 2018. 9

    2018

  38. [46]

    Training region-based object detectors with online hard ex- ample mining

    Abhinav Shrivastava, Abhinav Gupta, and Ross Girshick. Training region-based object detectors with online hard ex- ample mining. In CVPR, 2016. 9

    2016

  39. [47]

    Very deep convolu- tional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolu- tional networks for large-scale image recognition. In ICLR,

  40. [48]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In NeurIPS, 2017. 1

    2017

  41. [49]

    Graph attention networks

    Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adri- ana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks. In ICLR, 2018. 3

    2018

  42. [50]

    Understanding convolu- tion for semantic segmentation

    Panqu Wang, Pengfei Chen, Ye Yuan, Ding Liu, Zehua Huang, Xiaodi Hou, and Garrison Cottrell. Understanding convolu- tion for semantic segmentation. In WACV, 2018. 5

    2018

  43. [51]

    Non-local neural networks

    Xiaolong Wang, Ross Girshick, Abhinav Gupta, and Kaiming He. Non-local neural networks. In CVPR, 2018. 1, 2, 3, 5, 6, 7, 8, 9

    2018

  44. [52]

    Pay less attention with lightweight and dynamic convolutions

    Felix Wu, Angela Fan, Alexei Baevski, Yann N Dauphin, and Michael Auli. Pay less attention with lightweight and dynamic convolutions. In ICLR, 2019. 2

    2019

  45. [53]

    Aggregated residual transformations for deep neural networks

    Saining Xie, Ross Girshick, Piotr Doll´ar, Zhuowen Tu, and Kaiming He. Aggregated residual transformations for deep neural networks. In CVPR, 2017. 6

    2017

  46. [54]

    Learning deep structured multi-scale features using attention-gated crfs for contour prediction

    Dan Xu, Wanli Ouyang, Xavier Alameda-Pineda, Elisa Ricci, Xiaogang Wang, and Nicu Sebe. Learning deep structured multi-scale features using attention-gated crfs for contour prediction. In NeurIPS, 2017. 2

    2017

  47. [55]

    Denseaspp for semantic segmentation in street scenes

    Maoke Yang, Kun Yu, Chi Zhang, Zhiwei Li, and Kuiyuan Yang. Denseaspp for semantic segmentation in street scenes. In CVPR, 2018. 8

    2018

  48. [56]

    Bisenet: Bilateral segmentation network for real-time semantic segmentation

    Changqian Yu, Jingbo Wang, Chao Peng, Changxin Gao, Gang Yu, and Nong Sang. Bisenet: Bilateral segmentation network for real-time semantic segmentation. In ECCV, 2018. 9

    2018

  49. [57]

    Multi-scale context aggrega- tion by dilated convolutions

    Fisher Yu and Vladlen Koltun. Multi-scale context aggrega- tion by dilated convolutions. In ICLR, 2016. 2, 4, 6, 11

    2016

  50. [58]

    Ocnet: Object context network for scene parsing

    Yuhui Yuan and Jingdong Wang. Ocnet: Object context network for scene parsing. arXiv, 2018. 9

    2018

  51. [59]

    Dual graph convolutional network for semantic segmentation

    Li Zhang, Xiangtai Li, Anurag Arnab, Kuiyuan Yang, Yunhai Tong, and Philip HS Torr. Dual graph convolutional network for semantic segmentation. In BMVC, 2019. 5

    2019

  52. [60]

    Pyramid scene parsing network

    Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. In CVPR, 2017. 8, 9

    2017

  53. [61]

    Psanet: Point-wise spatial attention network for scene parsing

    Hengshuang Zhao, Yi Zhang, Shu Liu, Jianping Shi, Chen Change Loy, Dahua Lin, and Jiaya Jia. Psanet: Point-wise spatial attention network for scene parsing. In ECCV, 2018. 8, 9

    2018

  54. [62]

    Shuai Zheng, Sadeep Jayasumana, Bernardino Romera- Paredes, Vibhav Vineet, Zhizhong Su, Dalong Du, Chang Huang, and Philip H. S. Torr. Conditional random fields as recurrent neural networks. In ICCV, 2015. 2

    2015

  55. [63]

    De- formable convnets v2: More deformable, better results

    Xizhou Zhu, Han Hu, Stephen Lin, and Jifeng Dai. De- formable convnets v2: More deformable, better results. In CVPR, 2019. 2, 5, 6, 7

    2019

Pith tools

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