pith. machine review for the scientific record. sign in

arxiv: 2604.23399 · v2 · submitted 2026-04-25 · 💻 cs.CV

Recognition: unknown

Breaking the Resource Wall: Geometry-Guided Sequence Modeling for Efficient Semantic Segmentation

Sheng-Wei Chan , Hsin-Jui Pan , Chun-Po Shen , Chia-Min Lin , Yung-Che Wang , Jen-Shiun Chiang
Authors on Pith no claims yet

Pith reviewed 2026-05-08 08:37 UTC · model grok-4.3

classification 💻 cs.CV
keywords semantic segmentationstate space modelsgeometric guidanceefficient networksCityscapesADE20Kboundary preservationmamba architecture
0
0 comments X

The pith

DGM-Net achieves 82.3 percent mIoU on Cityscapes and 45.24 percent on ADE20K by guiding state space scans with centripetal flow fields and topological skeletons instead of scaling backbones or pretraining.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces DGM-Net to perform high-accuracy semantic segmentation while avoiding the usual demands for large pre-trained backbones or heavy compute. It replaces common context modules with a Directional Geometric Mamba operator that runs in linear time and adds a DGM-Module that pulls out centripetal flow fields plus topological skeletons to steer the scan order. These choices aim to keep boundary detail intact and deliver stable results even when training from scratch in under 30 thousand iterations on hardware limited to 8 GB VRAM and batch size 2. If the approach holds, it shows that structural geometric cues can substitute for raw model capacity in dense prediction tasks.

Core claim

DGM-Net introduces the G-Mamba operator as a linear-complexity alternative to ASPP and PPM, then augments it with the DGM-Module that extracts centripetal flow fields and topological skeletons to direct scanning; this combination yields 80.8 percent mIoU after 28k iterations, 82.3 percent mIoU on the Cityscapes test set, and 45.24 percent mIoU on ADE20K without large-scale pretraining or heavy backbone scaling, while remaining stable under batch size 2 on 8 GB VRAM.

What carries the argument

The DGM-Module extracts centripetal flow fields and topological skeletons to guide the scanning order inside the Directional Geometric Mamba (G-Mamba) operator.

If this is right

  • Reaches competitive mIoU scores after only 28k training iterations without pretraining.
  • Maintains performance when hardware forces batch size 2 on 8 GB VRAM.
  • Provides a linear-time replacement for heavier context modules such as ASPP and PPM.
  • Improves boundary detail through explicit geometric cues rather than larger model capacity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same geometric guidance pattern could be tested on other state-space or recurrent vision models for tasks like depth estimation or instance segmentation.
  • Lower hardware requirements may allow segmentation models to run on edge devices without sacrificing much accuracy.
  • Removing the need for large pretraining sets could make the method useful in domains where labeled data remain scarce.
  • Ablation studies that isolate flow-field versus skeleton contributions would clarify which geometric signal drives the gains.

Load-bearing premise

That feeding centripetal flow fields and topological skeletons into the G-Mamba scan will improve boundary preservation and overall accuracy without creating new instabilities or needing heavy extra tuning.

What would settle it

Training an identical backbone with the DGM-Module removed and measuring whether mIoU on boundary pixels or overall scores drops by more than 1-2 points on Cityscapes would test the claimed benefit of the geometric guidance.

Figures

Figures reproduced from arXiv: 2604.23399 by Chia-Min Lin, Chun-Po Shen, Hsin-Jui Pan, Jen-Shiun Chiang, Sheng-Wei Chan, Yung-Che Wang.

Figure 1
Figure 1. Figure 1: Overview of our proposed DGM-Net architecture. It follows an advanced encoder-decoder paradigm integrated with the Cascade view at source ↗
Figure 2
Figure 2. Figure 2: Visualization of geometric priors: (a) V -map showing potential fields, and (b) Curv-map capturing curvature anchors. layer. The fundamental difference between this mechanism and standard scanning is illustrated in view at source ↗
Figure 4
Figure 4. Figure 4: Computational complexity (GFLOPs) scaling across view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison on Cityscapes. Each row displays (from left to right): Ground Truth, V-map, D-Map, Flow Field, and view at source ↗
Figure 6
Figure 6. Figure 6: Visualization results on Cityscapes val set. view at source ↗
Figure 7
Figure 7. Figure 7: It shows qualitative ablation of the cascade design. view at source ↗
Figure 8
Figure 8. Figure 8: Even under a constrained setting (Batch Size = 2), DGM view at source ↗
read the original abstract

High-performance semantic segmentation has achieved significant progress in recent years, often driven by increasingly large backbones and higher computational budgets. While effective, such approaches introduce substantial computational overhead and limit accessibility under constrained hardware settings. In this paper, we propose DGM-Net (Directional Geometric Mamba Network), an efficient architecture that improves modeling capability through structural design rather than increasing model capacity. We introduce Directional Geometric Mamba (G-Mamba), a linear-complexity O(N) operator as an alternative to conventional context modeling modules such as ASPP and PPM. To further enhance structural awareness in state space model (SSM)-based modeling, we design the DGM-Module, which extracts centripetal flow fields and topological skeletons to guide the scanning process and improve boundary preservation. Without relying on large-scale pretraining or heavy backbone scaling, DGM-Net achieves 80.8% mIoU within 28k iterations, 82.3% mIoU on Cityscapes test set, and 45.24% mIoU on ADE20K. In addition, the model maintains stable performance under constrained hardware settings (e.g., batch size of 2 on 8GB VRAM), highlighting its efficiency and practicality. These results demonstrate that incorporating geometric guidance into SSM-based architectures provides an effective and resource-efficient direction for semantic segmentation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper introduces DGM-Net, an efficient semantic segmentation model that replaces conventional context modules (ASPP, PPM) with Directional Geometric Mamba (G-Mamba), a linear-complexity O(N) state-space operator. It further proposes the DGM-Module, which extracts centripetal flow fields and topological skeletons to guide the SSM scanning process and improve boundary preservation. Without large-scale pretraining or heavy backbone scaling, the model reports 80.8% mIoU after 28k iterations, 82.3% mIoU on the Cityscapes test set, and 45.24% mIoU on ADE20K, while remaining stable at batch size 2 on 8 GB VRAM.

Significance. If the results hold, the work shows that geometry-guided SSM scanning can deliver competitive semantic segmentation performance with substantially lower training cost and hardware requirements than backbone-scaling approaches. The manuscript supplies ablations, FLOPs comparisons, and cross-dataset results that directly support the headline numbers under the stated constraints, providing a concrete, resource-efficient alternative to current high-capacity designs.

minor comments (3)
  1. Abstract: the performance claims (80.8% mIoU in 28k iterations, 82.3% Cityscapes, 45.24% ADE20K) are stated without cross-references to the tables or sections that contain the corresponding ablations and baseline comparisons; adding these pointers would improve readability.
  2. §3 (DGM-Module description): the precise integration of centripetal flow fields and topological skeletons into the G-Mamba scan direction is described at a high level; a short pseudocode or explicit equation showing how the guidance modulates the state transition would eliminate ambiguity for readers wishing to re-implement the operator.
  3. Table captions and experimental protocol: error bars or standard deviations across multiple runs are not mentioned; including them (or stating that single-run results are reported) would strengthen the reproducibility claim.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive evaluation of our work, the recognition of DGM-Net's resource efficiency, and the recommendation for minor revision. We will incorporate any suggested minor changes in the revised manuscript.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper's central claims rest on an architectural design choice (DGM-Module extracting centripetal flow fields and topological skeletons to steer G-Mamba scanning) whose performance is validated through direct empirical results on Cityscapes and ADE20K under stated hardware constraints. No equations, parameters, or performance metrics are shown to reduce by construction to fitted inputs, self-definitions, or self-citation chains; the reported mIoU numbers and efficiency claims are presented as outcomes of the independent design rather than tautological restatements of the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The central claim rests on the unproven assumption that geometric flow and skeleton features meaningfully improve SSM scanning for segmentation; no free parameters are explicitly named in the abstract, but the 28k-iteration training schedule and choice of geometric extractors function as design choices whose justification is not supplied.

axioms (1)
  • domain assumption State-space models benefit from explicit geometric guidance during scanning for boundary-sensitive tasks
    Invoked when the DGM-Module is introduced to steer the G-Mamba operator.
invented entities (2)
  • Directional Geometric Mamba (G-Mamba) no independent evidence
    purpose: Linear-complexity O(N) operator replacing ASPP and PPM for context modeling
    New module proposed as the core of the architecture; no independent evidence outside the paper is provided.
  • DGM-Module no independent evidence
    purpose: Extracts centripetal flow fields and topological skeletons to guide scanning and preserve boundaries
    New component introduced to enhance structural awareness in SSMs; no external validation cited.

pith-pipeline@v0.9.0 · 5557 in / 1450 out tokens · 30163 ms · 2026-05-08T08:37:37.386773+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 7 canonical work pages · 4 internal anchors

  1. [1]

    Deep watershed transform for instance segmentation

    Min Bai and Raquel Urtasun. Deep watershed transform for instance segmentation. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 5221–5229, 2017. 2

    2017

  2. [2]

    Beit: Bert pre-training of image transformers

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. Beit: Bert pre-training of image transformers. InInternational Conference on Learning Representations, 2022. 1

    2022

  3. [3]

    The lov´asz-softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks

    Maxim Berman, Amal Rhouma Triki, and Matthew B Blaschko. The lov´asz-softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 4413–4421,

  4. [4]

    Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L Yuille. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolu- tion, and fully connected crfs.IEEE transactions on pattern analysis and machine intelligence, 40(4):834–848, 2017. 1

    2017

  5. [5]

    Rethinking Atrous Convolution for Semantic Image Segmentation

    Liang-Chieh Chen, George Papandreou, Florian Schroff, and Hartwig Adam. Rethinking atrous convolution for seman- tic image segmentation.arXiv preprint arXiv:1706.05587,

    work page internal anchor Pith review arXiv

  6. [6]

    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 Proceedings of the European conference on computer vision (ECCV), pages 801–818, 2018. 1, 2, 10

    2018

  7. [7]

    A semi-supervised boundary segmentation network for remote sensing images.Scientific Reports, 15(1):2007, 2025

    Yongdong Chen, Zaichun Yang, Liangji Zhang, and Weiwei Cai. A semi-supervised boundary segmentation network for remote sensing images.Scientific Reports, 15(1):2007, 2025. 2

    2007

  8. [8]

    Masked-attention mask transformer for universal image segmentation

    Bowen Cheng, Ishan Misra, Alexander G Schwing, Alexan- der Kirillov, and Rohit Girdhar. Masked-attention mask transformer for universal image segmentation. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1290–1299, 2022. 2, 8

    2022

  9. [9]

    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. InProceed- ings of the IEEE conference on computer vision and pattern recognition, pages 3213–3223, 2016. 7

    2016

  10. [10]

    Deformable convolutional networks

    Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. Deformable convolutional networks. InProceedings of the IEEE international confer- ence on computer vision, pages 764–773, 2017. 6

    2017

  11. [11]

    Centripetalnet: Pursuing high- quality keypoint-based object detection

    Zhiwei Dong, Guoxuan Li, Yue Liao, Fei Wang, Pengju Ren, and Chen Qian. Centripetalnet: Pursuing high- quality keypoint-based object detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10519–10528, 2020. 2

    2020

  12. [12]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy et al. An image is worth 16x16 words: Transformers for image recognition at scale. InICLR, 2021. 1

    2021

  13. [13]

    Ecmnet: Lightweight semantic segmentation with efficient cnn-mamba network

    Feixiang Du, Shengkun Wu, Xiang Wang, Aoxue Ding, Zhongliang Wang, and Joel CM Than. Ecmnet: Lightweight semantic segmentation with efficient cnn-mamba network. Science Progress, 109(1):00368504261419245, 2026. 2

    2026

  14. [14]

    Dual attention network for scene seg- mentation

    Jun Fu, Jing Liu, Haijie Tian, Yong Li, Yongjun Bao, Zhiwei Fang, and Hanqing Lu. Dual attention network for scene seg- mentation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3146–3154,

  15. [15]

    Mamba: Linear-Time Sequence Modeling with Selective State Spaces

    Albert Gu and Tri Dao. Mamba: Linear-time sequence modeling with selective state spaces.arXiv preprint arXiv:2312.00752, 2023. 1, 2

    work page internal anchor Pith review arXiv 2023

  16. [16]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 3

    2016

  17. [17]

    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. InProceedings of the IEEE/CVF international conference on computer vision, pages 603–612, 2019. 2

    2019

  18. [18]

    Alignseg: Feature- aligned segmentation networks

    Zilong Huang, Yunchao Wei, Xinggang Wang, Wenyu Liu, Thomas S Huang, and Humphrey Shi. Alignseg: Feature- aligned segmentation networks. InIEEE Transactions on Pattern Analysis and Machine Intelligence, 2021. 3

    2021

  19. [19]

    Spatial transformer networks

    Max Jaderberg, Karen Simonyan, Andrew Zisserman, and Koray Kavukcuoglu. Spatial transformer networks. InAd- vances in neural information processing systems, volume 28,

  20. [20]

    Global and local mamba network for multi- modality medical image super-resolution.Pattern Recogni- tion, page 112888, 2025

    Zexin Ji, Beiji Zou, Xiaoyan Kui, Sbastien Thureau, and Su Ruan. Global and local mamba network for multi- modality medical image super-resolution.Pattern Recogni- tion, page 112888, 2025. 2

    2025

  21. [21]

    Afrda: Attentive feature refinement for domain adaptive se- mantic segmentation.IEEE Robotics and Automation Let- ters, 2025

    Md Al-Masrur Khan, Durgakant Pushp, and Lantao Liu. Afrda: Attentive feature refinement for domain adaptive se- mantic segmentation.IEEE Robotics and Automation Let- ters, 2025. 3

    2025

  22. [22]

    Pointrend: Image segmentation as rendering

    Alexander Kirillov, Yuxin Wu, Kaiming He, and Ross Gir- shick. Pointrend: Image segmentation as rendering. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9769–9778, 2020. 2

    2020

  23. [23]

    Semantic flow for fast and accurate scene parsing

    Xiangtai Li, Ansheng You, Zhen Zhu, Houlong Zhao, Ming Yang, Kuiyuan Yang, and Yunhai Tong. Semantic flow for fast and accurate scene parsing. InEuropean Conference on Computer Vision, pages 775–793. Springer, 2020. 3

    2020

  24. [24]

    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 European conference on computer vision, pages 740–755. Springer, 2014. 7

    2014

  25. [25]

    arXiv preprint arXiv:2401.10166 , year=

    Yue Liu, Yunjie Tian, Yuzhi Zhao, Hongtian Yu, Lingxi Xie, Yaowei Wang, Qixiang Ye, and Yu Qiao. Vmamba: Visual state space model.arXiv preprint arXiv:2401.10166, 2024. 2

    work page arXiv 2024

  26. [26]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 8

    2021

  27. [27]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yunchao Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 10012–10022, 2021. 1

    2021

  28. [28]

    A convnet for the 2020s

    Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feicht- enhofer, Trevor Darrell, and Saining Xie. A convnet for the 2020s. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 11976–11986,

  29. [29]

    Fully convolutional networks for semantic segmentation

    Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. InPro- ceedings of the IEEE conference on computer vision and pat- tern recognition, pages 3431–3440, 2015. 1, 2

    2015

  30. [30]

    Decoupled Weight Decay Regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017. 7

    work page internal anchor Pith review arXiv 2017

  31. [31]

    Attention-based conditioning methods for ex- ternal knowledge integration

    Katerina Margatina, Christos Baziotis, and Alexandros Potamianos. Attention-based conditioning methods for ex- ternal knowledge integration. InProceedings of the 57th an- nual meeting of the association for computational linguis- tics, pages 3944–3951, 2019. 2

    2019

  32. [32]

    Vcmamba: Bridging convolutions with multi-directional mamba for efficient visual representation

    Mustafa Munir, Alex Zhang, and Radu Marculescu. Vcmamba: Bridging convolutions with multi-directional mamba for efficient visual representation. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 3037–3046, 2025. 2

    2025

  33. [33]

    The mapillary vistas dataset for semantic understanding of street scenes

    Gerhard Neuhold, Tobias Ollmann, Samuel Rota Bulo, and Peter Kontschieder. The mapillary vistas dataset for semantic understanding of street scenes. InProceedings of the IEEE international conference on computer vision, pages 4990– 4999, 2017. 7

    2017

  34. [34]

    U- net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- net: Convolutional networks for biomedical image segmen- tation. InInternational Conference on Medical image com- puting and computer-assisted intervention, pages 234–241. Springer, 2015. 2

    2015

  35. [35]

    Imagenet large scale visual recognition challenge.International journal of computer vision, 115(3):211–252, 2015

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, et al. Imagenet large scale visual recognition challenge.International journal of computer vision, 115(3):211–252, 2015. 7

    2015

  36. [36]

    Visual state space models with spiral selective scan for referring remote-sensing image segmentation.Geo- spatial Information Science, pages 1–19, 2026

    Weihao Shen, Ailong Ma, Zhuo Zheng, Junjue Wang, and Yanfei Zhong. Visual state space models with spiral selective scan for referring remote-sensing image segmentation.Geo- spatial Information Science, pages 1–19, 2026. 2

    2026

  37. [37]

    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. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 761–769,

  38. [38]

    Gated convolutional neural network for semantic segmentation in high-resolution images

    Towaki Takikawa, David Acuna, Varun Jampani, and Sanja Fidler. Gated convolutional neural network for semantic segmentation in high-resolution images. InProceedings of the IEEE/CVF international conference on computer vision, pages 5287–5296, 2019. 2, 8

    2019

  39. [39]

    Training data-efficient image transform- ers & distillation through attention

    Hugo Touvron et al. Training data-efficient image transform- ers & distillation through attention. InICML, 2021. 1

    2021

  40. [40]

    Attention is all you need

    Ashish Vaswani et al. Attention is all you need. InAdvances in neural information processing systems, 2017. 1

    2017

  41. [41]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017. 2

    2017

  42. [42]

    Internimage: Exploring large-scale vision founda- tion models with deformable convolutions

    Jice Wang, Jifeng Dai, Zhe Chen, Lewei Huang, Zhiqi Li, Xizhou Zhu, Xiaowei Hu, Tong Lu, Lewei Lu, Hongsheng Li, et al. Internimage: Exploring large-scale vision founda- tion models with deformable convolutions. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14408–14419, 2023. 1

    2023

  43. [43]

    Dfam-detr: De- formable feature based attention mechanism detr on slender object detection.IEICE TRANSACTIONS on Information and Systems, 106(3):401–409, 2023

    Feng Wen, Mei Wang, and Xiaojie Hu. Dfam-detr: De- formable feature based attention mechanism detr on slender object detection.IEICE TRANSACTIONS on Information and Systems, 106(3):401–409, 2023. 3

    2023

  44. [44]

    Segformer: Simple and efficient design for semantic segmentation with transform- ers.Advances in neural information processing systems, 34:12077–12090, 2021

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M Alvarez, and Ping Luo. Segformer: Simple and efficient design for semantic segmentation with transform- ers.Advances in neural information processing systems, 34:12077–12090, 2021. 1, 2, 8

    2021

  45. [45]

    Deeperlab: Single-shot image parser.arXiv preprint arXiv:1902.05093, 2019

    Tien-Ju Yang, Maxwell D Collins, Yukun Zhu, Jyh-Jing Hwang, Ting Liu, Xiao Zhang, Vivienne Sze, George Pa- pandreou, and Liang-Chieh Chen. Deeperlab: Single-shot image parser.arXiv preprint arXiv:1902.05093, 2019. 2

    work page arXiv 1902

  46. [46]

    Context prior for scene segmentation

    Changqian Yu, Jingbo Wang, Changxin Gao, Gang Yu, Chunhua Shen, and Nong Sang. Context prior for scene segmentation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12416– 12425, 2020. 2

    2020

  47. [47]

    Multi-Scale Context Aggregation by Dilated Convolutions

    Fisher Yu and Vladlen Koltun. Multi-scale context aggregation by dilated convolutions.arXiv preprint arXiv:1511.07122, 2015. 3

    work page Pith review arXiv 2015

  48. [48]

    Object- contextual representations for semantic segmentation

    Yuhui Yuan, Xilin Chen, and Jingdong Wang. Object- contextual representations for semantic segmentation. In European conference on computer vision, pages 173–190. Springer, 2020. 8, 9

    2020

  49. [49]

    Compact generalized non-local network.Ad- vances in neural information processing systems, 31, 2018

    Kaiyu Yue, Ming Sun, Yuchen Yuan, Feng Zhou, Errui Ding, and Fuxin Xu. Compact generalized non-local network.Ad- vances in neural information processing systems, 31, 2018. 2

    2018

  50. [50]

    Con- text encoding for semantic segmentation

    Hang Zhang, Kristin Dana, Jianping Shi, Zhongyue Zhang, Xiaogang Wang, Ambrish Tyagi, and Amit Agrawal. Con- text encoding for semantic segmentation. InProceedings of the IEEE conference on computer vision and pattern recog- nition, pages 7151–7160, 2018. 2

    2018

  51. [51]

    Dual-path knowledge-augmented contrastive alignment network for spatially resolved transcriptomics

    Wei Zhang, Jiajun Chu, Xinci Liu, Chen Tong, and Xinyue Li. Dual-path knowledge-augmented contrastive alignment network for spatially resolved transcriptomics. InProceed- ings of the AAAI Conference on Artificial Intelligence, vol- ume 40, pages 12807–12815, 2026. 3

    2026

  52. [52]

    Pyramid scene parsing network

    Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2881–2890, 2017. 1, 2, 10

    2017

  53. [53]

    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. InProceedings of the European conference on computer vision (ECCV), pages 267–283, 2018. 8

    2018

  54. [54]

    Rsmamba: Remote sensing image classifi- cation with state space model.IEEE Geoscience and Remote Sensing Letters, 2024

    Keyan Zhao et al. Rsmamba: Remote sensing image classifi- cation with state space model.IEEE Geoscience and Remote Sensing Letters, 2024. 2

    2024

  55. [55]

    Rethinking semantic segmen- tation from a sequence-to-sequence perspective with trans- formers

    Sixiao Zheng, Jiachen Lu, Hengshuang Zhao, Xiatian Zhu, Zekun Luo, Yabiao Wang, Yanwei Fu, Jianfeng Feng, Tao Xiang, Philip HS Torr, et al. Rethinking semantic segmen- tation from a sequence-to-sequence perspective with trans- formers. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6881–6890,

  56. [56]

    Scene parsing through ade20k dataset

    Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. Scene parsing through ade20k dataset. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 633–641,

  57. [57]

    Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model

    Lianghui Zhu, Bencheng Liao, Qian Zhang, Xinlong Wang, Wenyu Liu, and Xinggang Wang. Vision mamba: Efficient visual representation learning with bidirectional state space model.arXiv preprint arXiv:2401.09417, 2024. 2, 8

    work page internal anchor Pith review arXiv 2024

  58. [58]

    Asymmetric non-local neural net- works for semantic segmentation

    Zhen Zhu, Mengde prescribed Xu, Song Bai, Tengteng Huang, and Xiang Bai. Asymmetric non-local neural net- works for semantic segmentation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 593–602, 2019. 2

    2019