Pith. sign in

REVIEW 4 major objections 6 minor 27 references

SCANet: Split Coordinate Attention Network for Building Footprint Extraction

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A split coordinate attention module that pools each channel along x and y axes improves building footprint extraction on two public datasets.

desk verdict A coherent attention-module hybrid with consistent internal gains, but the SOTA claim needs error bars and clarified backbones before it is trustworthy. read the letter →

arxiv 2507.20809 v1 pith:4O3GJNVC submitted 2025-07-28 cs.CV

Chunshi Wang , Bin Zhao , Shuxue Ding This is my paper · ORCID
classification cs.CV
keywords buildingfootprintextractionimagesegmentationremotesensingattentionmechanismcoordinateconvolutionalneuralnetworkUNet++split
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

Building footprint extraction from remote sensing images is usually done with convolutional networks whose attention modules either summarise each channel globally or lose positional detail. This paper claims that a new plug-and-play module, Split Coordinate Attention (SCA), fixes that by pooling each channel separately along the horizontal and vertical axes and then splitting each feature group into weighted branches. Inserted into a ResNet or ResNeSt encoder and paired with a UNet++ decoder, the resulting SCANet reaches 91.61% IoU on the WHU Building Dataset and 75.49% on the Massachusetts Building Dataset, the best among the compared methods. The authors argue this shows that a targeted CNN attention design can outperform transformer-based extraction models while using no more parameters.

What carries the argument

The load-bearing object is the Split Coordinate Attention (SCA) module. It decomposes the global pooling of split attention into two pooling kernels, $(H,1)$ along the horizontal direction and $(1,W)$ along the vertical direction, so each channel is encoded with its row and column context rather than a single scalar. The two feature maps are concatenated, reduced by a shared dense layer with batch normalisation and ReLU, split back into horizontal and vertical branches, and each branch passes through its own dense layer and sigmoid. A cardinal group's output is the input times the sum over splits of the product of the two directional weights; this product is the operation that lets the module weight long-range spatial interactions per channel.

What would settle it

Run SCANet and BCTNet on the WHU Building Dataset with five random seeds each, record mean and standard deviation of IoU; if the best-baseline margin overlaps zero or the ordering flips, the claimed state-of-the-art result is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that replacing the global average pooling used in split attention with two directional pooling operations makes channel attention retain positional information. For each cardinal group, SCA pools along (H,1) and (1,W), concatenates the two direction-aware descriptors, passes them through a shared dense layer, splits the result back into height and width branches, and applies per-branch dense connections with sigmoid gating. Each split then votes with the product of its horizontal and vertical weights, so the network can emphasise spatially remote but semantically related pixels. With this module inside a CNN encoder, SCANet reports state-of-the-art intersection-over-union on both benchmark datasets and, in the ablation study, improves over both split attention and coordinate attention at equal or smaller parameter counts.

Load-bearing premise

The reported advantage rests on the assumption that the 0.45-0.46 percentage-point IoU margins over the strongest baselines are stable, reproducible effects rather than noise from a single training run.

Editorial extensions

If this is right

  • Plugging SCA into an existing CNN encoder can raise building footprint IoU without increasing parameter count, so it may be a drop-in upgrade for deployed segmentation models.
  • Because SCA beats transformer-based baselines, the paper implies that long-range context in remote sensing can be obtained with cheap directional pooling rather than self-attention.
  • The module's parameter parity with split attention and coordinate attention means its gains come from where the pooling is applied, not from added capacity.
  • UNet++ as decoder is needed to realise the full gain: with a plain UNet decoder the same module reaches 90.41% IoU on WHU, still above baselines but below the reported peak.

Reading between the lines

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

  • The diagonal-similarity visualisation suggests SCA enlarges the effective receptive field of the final encoder stage; a direct measurement of receptive field size across the compared backbones would test that explanation.
  • The same two-axis pooling design could be applied to other dense prediction tasks, such as road or water mapping, where buildings occupy a small fraction of the image; that test is not in the paper.
  • Because the Massachusetts dataset has lower resolution and noisier labels than WHU, the consistent margin there hints the module is robust to label noise, an idea the paper does not pursue.
  • The reported gaps are single runs; multi-seed evaluation would tell whether the improvement is stable enough for practical deployment.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes Split Coordinate Attention (SCA), a plug-and-play attention module that decomposes the global pooling used in split attention into horizontal and vertical pooled descriptors, applies per-group dense transformations, and combines horizontal and vertical soft-attention weights to reweight cardinal-group features. Inserting SCA into ResNet/ResNeSt backbones and decoding with UNet++ yields SCANet. The authors report state-of-the-art IoU values of 91.61% on the WHU Building Dataset and 75.49% on the Massachusetts Building Dataset, together with ablations, complexity comparisons, Grad-CAM++ visualizations, and a diagonal pixel-similarity analysis. The module equations in Section 2.2 are coherent, and the parameter analysis in Table 4 indicates that the gains are not obtained simply by enlarging the model. The central weakness is empirical: the state-of-the-art claim rests on single-run IoU differences of 0.45-0.46 percentage points over strong baselines, with no error bars, no repeated seeds, and no evidence that the uniform training protocol is appropriate for transformer-based baselines.

Significance. If the reported numbers are reproducible and the comparisons are fair, the paper would provide a useful lightweight attention module for building footprint extraction: it is simple, adds no parameters relative to ResNeSt in the main configuration, is evaluated on two standard public benchmarks, and the code is released. The interpretability analyses (Grad-CAM++ and diagonal similarity) are a positive complement, and the complexity analysis is a useful check that the reported gains are not simply a consequence of larger models. However, the quantitative improvement over BCTNet is modest (0.45-0.46 IoU), and the paper's principal contribution is empirical. Without statistical reliability and a clearly fair baseline protocol, the claimed state-of-the-art performance is not established. The module itself is a reasonable combination of coordinate attention and split attention, but its practical value depends on resolving these empirical concerns.

major comments (4)
  1. [Section 3.3, Tables 1 and 2] The abstract and Section 3.3 claim state-of-the-art performance based on single-run IoU differences of 0.46% (WHU) and 0.45% (MASS) over BCTNet, but no standard deviations, confidence intervals, or multiple-seed experiments are reported. On these benchmark splits, seed-to-seed IoU variation for segmentation models is commonly in the 0.3-0.8 range, so the observed margins are within plausible run-to-run noise. Please report means and standard deviations over at least three independent runs for SCANet and the nearest baselines, and state whether the ordering is consistent across seeds. This is load-bearing because the central claim of the paper is empirical superiority.
  2. [Sections 3.2 and 2.3] The comparison may be unfair to transformer-based baselines. Section 3.2 states that 'consistent hyperparameter settings' are used for all experiments, including batch size 4 on MASS, AdamW with learning rate 1e-3, and IoU-based learning-rate halving. Methods such as BCTNet, TransFuse-L, and Swin generally require different learning rates, warmup schedules, and training lengths; a one-size-fits-all recipe can systematically depress their reported IoU. Please either run each baseline with its official or recommended training configuration, or provide evidence that the chosen protocol is near-optimal for every baseline. In addition, Section 2.3 does not specify where SCA is inserted into the backbone: which stages, which residual blocks, and how many SCA blocks are used. This information is needed to reproduce SCANet.
  3. [Table 4] Table 4 lists 'ResNet-14' and 'ResNet-26' as backbones. Standard ResNet depths are 18, 34, 50, 101, and 152; ResNet-14 and ResNet-26 are non-standard and their block configurations are not described. Without stating the number of blocks per stage, stem, stride pattern, and how 'ResNet-14/26' relate to standard ResNets, the parameter counts and IoU values in this table cannot be verified or compared. Please clarify these configurations and, if they are custom variants, define them precisely.
  4. [Section 3.4, Table 3] The ablation table mixes changes in the attention module and the decoder, and the parameter counts are not self-explanatory: Baseline UNet++ has 68.0M, +SA has 73.2M, +CA has 79.2M, +SCA UNet++ has 73.2M, and +SCA UNet has 55.3M. If +SCA and +SA have identical parameter counts, the table should state explicitly what backbone and insertion points are used for each row. If the baseline row uses a different backbone or insertion configuration than the +SCA row, the comparison does not isolate the effect of the module. Please report the backbone, insertion locations, and decoder for every row, and add a row with SCA inserted into the same backbone and decoder as the baseline.
minor comments (6)
  1. [Section 2.2, Eq. (5)] The symbols S_h_c and S_w_c are described as 'Dense connections' but are not formally defined; please specify the input/output dimensions, the number of layers, and the growth rate of these dense connections.
  2. [Figure 1] Figure 1 contains labels 'PreQuantization Stage 1' and 'PostQuantization Stage N' that are not defined anywhere in the text; please define them or remove them.
  3. [Throughout] There are several typographical and grammar issues: 'featuremap Group' in Section 2.1, 'the c-th channel is calculated' in Section 2.2, and 'two spatial range of pooling kernels' in the abstract. These should be corrected.
  4. [Section 2.2, Eq. (4)] Equation (4) uses [·, ·] for joining elements, and the phrase 'activation function applied after convolution' for G_c^i is vague; please specify whether G_c^i is a convolutional block with batch norm and ReLU, and give the kernel size.
  5. [References and code] The code URL is given but no commit hash is provided, so the exact configuration cannot be pinned. Reference [12] is cited only as 'arXiv preprint (2017)' with no identifier; please add the arXiv number.
  6. [Table 4 versus Tables 1-2] Table 4 reports IoU for Swin UNETR and Swin-B, but these methods do not appear in Tables 1-2; the text does not describe the experimental setup for these additional comparisons. Please clarify whether these numbers come from the same protocol and where the corresponding details are reported.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SCA is an explicit composition of cited prior attention designs and all headline numbers come from held-out public test splits.

full rationale

The paper's central construction, Split Coordinate Attention (SCA), is built openly from prior components: Equations (2)-(3) apply the horizontal/vertical pooling decomposition of Coordinate Attention [6], Equation (4) explicitly follows 'the coordinate attention generation designed in [6]', and the grouped split weighting follows ResNeSt [23]. No equation in Section 2 defines a predicted quantity in terms of the target result; the module is an architectural composition, not a derivation. The performance claims are empirical evaluations on the public WHU and Massachusetts test splits, with no reported test-set fitting, so the IoU numbers are not forced by construction. The only self-overlap is that one co-author (Bin Zhao) appears on the ConvBNet baseline paper [21]; that baseline is an external benchmark entry, and the comparison does not reduce to the cited work. Table 3 ablates SA, CA, and SCA under the same protocol, and the claimed advantage is an experimental measurement rather than an identity. The absence of error bars and the unusual ResNet-14/ResNet-26 labels in Table 4 are statistical and comparability concerns, not circularity. No fitted parameter is renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work. The derivation chain is therefore self-contained with respect to the specific circularity patterns under review.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

The central claim rests on a few undisclosed hyperparameters (K, R, c') and the assumption that standard backbones and training recipes apply uniformly. No entirely new physical or mathematical entity is introduced beyond the SCA module itself.

free parameters (3)
  • K (number of cardinal groups)
    Hyperparameter controlling number of feature groups; value not reported in the paper.
  • R (number of splits per cardinal group)
    Hyperparameter controlling number of splits; value not reported.
  • Intermediate channel count c' in the split dense module
    Unspecified dimension used in the Dense layers after concatenation.
assumptions (3)
  • domain assumption Standard ResNet and ResNeSt backbones as defined in He et al. (2016) and Zhang et al. (2022) are valid feature extractors
    The paper builds SCANet on these backbones and assumes their pre-defined architectures are correctly loaded and used.
  • domain assumption The public WHU and Massachusetts datasets are used with standard splits and preprocessing
    Section 3.1 describes the data but does not verify split consistency or label quality beyond citing the sources.
  • domain assumption The training recipe (BCE plus Dice loss, AdamW, LR schedule) transfers fairly to all compared methods
    Section 3.2 states consistent hyperparameters, implying optimal settings for all, which is an untested assumption.
invented entities (1)
  • Split Coordinate Attention (SCA) module
    purpose: New attention mechanism to capture spatial and channel interactions for segmentation
    Effectiveness is demonstrated only on the two datasets inside this paper; no external validation or theoretical guarantee exists beyond the reported experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SCANet: Split Coordinate Attention Network for Building Footprint Extraction." pith.science (2026). https://pith.science/paper/4O3GJNVC

@misc{pith2026250720809,
  author       = {Pith},
  title        = {Pith review of: SCANet: Split Coordinate Attention Network for Building Footprint Extraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4O3GJNVC}},
  note         = {Machine review of arXiv:2507.20809}
}
read the original abstract

Building footprint extraction holds immense significance in remote sensing image analysis and has great value in urban planning, land use, environmental protection and disaster assessment. Despite the progress made by conventional and deep learning approaches in this field, they continue to encounter significant challenges. This paper introduces a novel plug-and-play attention module, Split Coordinate Attention (SCA), which ingeniously captures spatially remote interactions by employing two spatial range of pooling kernels, strategically encoding each channel along x and y planes, and separately performs a series of split operations for each feature group, thus enabling more efficient semantic feature extraction. By inserting into a 2D CNN to form an effective SCANet, our SCANet outperforms recent SOTA methods on the public Wuhan University (WHU) Building Dataset and Massachusetts Building Dataset in terms of various metrics. Particularly SCANet achieves the best IoU, 91.61% and 75.49% for the two datasets. Our code is available at https://github.com/AiEson/SCANet

Figures

Figures reproduced from arXiv: 2507.20809 by the authors.

Figure 1
Figure 1. The overall framework of the proposed method. For easy visualization, [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The figure shows the X,Y Avg Pooling module (a) and the Split Dense [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Qualitative results on the WHU Dataset. (a) Image. (b) GT. (c) SCANet. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Qualitative results on the MASS Dataset. (a) Image. (b) GT. (c) SCANet. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Grad-CAM++ visualization of the proposed SCANet and other methods [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: The diagonal similarity of different encoders in the final stage. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 17 canonical work pages

  1. [1]

    In: 2018 IEEE winter conference on applications of computer vision (WACV)

    Chattopadhay, A., Sarkar, A., Howlader, P., Balasubramanian, V.N.: Grad- cam++: Generalized gradient-based visual explanations for deep convolutional networks. In: 2018 IEEE winter conference on applications of computer vision (WACV). pp. 839–847. IEEE (2018)

    2018

  2. [2]

    Remote Sensing13(21), 4441 (2021)

    Chen, K., Zou, Z., Shi, Z.: Building extraction from remote sensing images with sparse token transformers. Remote Sensing13(21), 4441 (2021)

    work page 2021

  3. [3]

    In: Proceedings of the European conference on computer vision (ECCV)

    Chen, L.C., Zhu, Y., Papandreou, G., Schroff, F., Adam, H.: Encoder-decoder with atrous separable convolution for semantic image segmentation. In: Proceedings of the European conference on computer vision (ECCV). pp. 801–818 (2018)

    2018

  4. [4]

    Remote Sensing Letters3(2), 151– 159 (2012)

    Cui, S., Yan, Q., Reinartz, P.: Complex building description and extraction based on hough transformation and cycle detection. Remote Sensing Letters3(2), 151– 159 (2012)

    work page 2012

  5. [5]

    He,K.,Zhang,X.,Ren,S.,Sun,J.:Deepresiduallearningforimagerecognition.In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016)

    2016

  6. [6]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Hou, Q., Zhou, D., Feng, J.: Coordinate attention for efficient mobile network de- sign. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 13713–13722 (2021)

    work page 2021

  7. [7]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 7132–7141 (2018)

    2018

  8. [8]

    In: Proceedings of the IEEE/CVF interna- tional conference on computer vision

    Huang, Z., Wang, X., Huang, L., Huang, C., Wei, Y., Liu, W.: Ccnet: Criss-cross attention for semantic segmentation. In: Proceedings of the IEEE/CVF interna- tional conference on computer vision. pp. 603–612 (2019) SCANet: Split Coordinate Attention Network 13

    work page 2019

Show all 27 references
  1. [9]

    IEEE Transactions on Geoscience and Remote Sensing57(1), 574–586 (2018)

    Ji, S., Wei, S., Lu, M.: Fully convolutional networks for multisource building ex- traction from an open aerial and satellite imagery data set. IEEE Transactions on Geoscience and Remote Sensing57(1), 574–586 (2018)

    2018

  2. [10]

    In: Proceedings of the eighth ACM symposium on Solid modeling and applications

    Laycock, R.G., Day, A.: Automatically generating large urban environments based on the footprint data of buildings. In: Proceedings of the eighth ACM symposium on Solid modeling and applications. pp. 346–351 (2003)

    2003

  3. [11]

    Li,X.,Wang,W.,Hu,X.,Yang,J.:Selectivekernelnetworks.In:Proceedingsofthe IEEE/CVF conference on computer vision and pattern recognition. pp. 510–519 (2019)

    2019

  4. [12]

    Loshchilov, I., Hutter, F.: Fixing weight decay regularization in adam (2017)

    2017

  5. [13]

    Mnih, V.: Machine Learning for Aerial Image Labeling. Ph.D. thesis, University of Toronto (2013)

    2013

  6. [14]

    object-based classification of urban land cover extraction using high spatial reso- lution imagery

    Myint, S.W., Gober, P., Brazel, A., Grossman-Clarke, S., Weng, Q.: Per-pixel vs. object-based classification of urban land cover extraction using high spatial reso- lution imagery. Remote sensing of environment115(5), 1145–1161 (2011)

    2011

  7. [15]

    ISPRS Journal of Photogrammetry and Remote Sensing 62(1), 43–63 (2007)

    Sohn, G., Dowman, I.: Data fusion of high-resolution satellite imagery and lidar data for automatic building extraction. ISPRS Journal of Photogrammetry and Remote Sensing 62(1), 43–63 (2007)

    2007

  8. [16]

    arXiv preprint arXiv:1904.04514 (2019)

    Sun, K., Zhao, Y., Jiang, B., Cheng, T., Xiao, B., Liu, D., Mu, Y., Wang, X., Liu, W., Wang, J.: High-resolution representations for labeling pixels and regions. arXiv preprint arXiv:1904.04514 (2019)

    2019 arXiv

  9. [17]

    International Archives of Photogrammetry and Remote Sensing33(B3/2; PART 3), 958–964 (2000)

    Wang, Z., Schenk, T.: Building extraction and reconstruction from lidar data. International Archives of Photogrammetry and Remote Sensing33(B3/2; PART 3), 958–964 (2000)

    2000

  10. [18]

    In: Proceedings of the European conference on computer vision (ECCV)

    Woo, S., Park, J., Lee, J.Y., Kweon, I.S.: Cbam: Convolutional block attention module. In: Proceedings of the European conference on computer vision (ECCV). pp. 3–19 (2018)

    2018

  11. [19]

    Advances in Neural Information Processing Systems34, 12077–12090 (2021)

    Xie, E., Wang, W., Yu, Z., Anandkumar, A., Alvarez, J.M., Luo, P.: Segformer: Simple and efficient design for semantic segmentation with transformers. Advances in Neural Information Processing Systems34, 12077–12090 (2021)

    2021

  12. [20]

    IEEE Transactions on Geoscience and Remote Sensing (2023)

    Xu, L., Li, Y., Xu, J., Zhang, Y., Guo, L.: Bctnet: Bi-branch cross-fusion trans- former for building footprint extraction. IEEE Transactions on Geoscience and Remote Sensing (2023)

    2023

  13. [21]

    IEEE Geoscience and Remote Sensing Letters 20, 1–5 (2023)

    Yu, T., Tang, P., Zhao, B., Bai, S., Gou, P., Liao, J., Jin, C.: Convbnet: A convo- lutional network for building footprint extraction. IEEE Geoscience and Remote Sensing Letters 20, 1–5 (2023)

    2023

  14. [22]

    IEEE Geoscience and Remote Sensing Letters19, 1–5 (2021)

    Yu, Y., Liu, C., Gao, J., Jin, S., Jiang, X., Jiang, M., Zhang, H., Zhang, Y.: Building extraction from remote sensing imagery with a high-resolution capsule network. IEEE Geoscience and Remote Sensing Letters19, 1–5 (2021)

    2021

  15. [23]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Zhang, H., Wu, C., Zhang, Z., Zhu, Y., Lin, H., Zhang, Z., Sun, Y., He, T., Mueller, J., Manmatha, R., et al.: Resnest: Split-attention networks. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 2736– 2746 (2022)

    2022

  16. [24]

    arXiv preprint arXiv:2301.01146 (2023)

    Zhang, J., Li, X., Li, J., Liu, L., Xue, Z., Zhang, B., Jiang, Z., Huang, T., Wang, Y., Wang, C.: Rethinking mobile block for efficient neural models. arXiv preprint arXiv:2301.01146 (2023)

    2023 arXiv

  17. [25]

    In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part I 24

    Zhang, Y., Liu, H., Hu, Q.: Transfuse: Fusing transformers and cnns for medi- cal image segmentation. In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part ...

    2021

  18. [26]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhao, H., Shi, J., Qi, X., Wang, X., Jia, J.: Pyramid scene parsing network. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 2881–2890 (2017)

    2017

  19. [27]

    Zhou, Z., Rahman Siddiquee, M.M., Tajbakhsh, N., Liang, J.: Unet++: A nested u-net architecture for medical image segmentation. In: Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th In- ternational Workshop, DLMIA 2018, and 8th ...

    2018

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.