Pith. sign in

REVIEW 3 major objections 6 minor 40 references

BAFPN: Bi directional alignment of features to improve localization accuracy

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A feature pyramid that first aligns features bottom-up at the global scale, then aligns them semantically top-down, raises high-IoU detection accuracy on rotated objects, with AP75 up 1.68 points on DOTAv1.5.

desk verdict Solid incremental FPN variant; the global-alignment claim is softer than the paper admits, but the empirical gains warrant referee time. read the letter →

arxiv 2412.01859 v1 pith:6I3DSTDN submitted 2024-12-01 cs.CV

Li Jiakun , Wang Qingqing , Dong Hongbin , Li Kexin This is my paper · ORCID
classification cs.CV
keywords featurepyramidnetworkspatialalignmentsemanticorientedobjectdetectionremotesensingdeformableconvolutionDOTAv1.5localization
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

The paper targets a specific failure mode of feature pyramid networks in remote-sensing object detection: deep features can be spatially shifted and deformed relative to the original image, so top-down fusion mixes them into wrong places and hurts high-precision localization. BAFPN adds a bottom-up pass before the usual top-down fusion, where the Spatial Feature Alignment Module aligns each deeper feature map to the immediately shallower one, so position and edge corrections cascade from the finest level upward and act at the global scale. A second module, SEAM, then masks deep features at channel and pixel level during top-down fusion to shrink semantic gaps, and GALM replaces 1x1 lateral connections with grouped aggregation to avoid channel loss. On DOTAv1.5, swapping FPN for BAFPN in Oriented R-CNN raises AP75 by 1.68 points, AP50 by 1.45 points, and mAP by 1.34 points, and similar or larger gains appear in other rotated detectors, especially anchor-free ones. The paper's thesis is that global spatial alignment, not just richer fusion, is what carries these localization gains.

What carries the argument

The load-bearing object is the bottom-up Spatial Feature Alignment Module (SPAM) cascade. Within SPAM, a spatial-to-depth downsampler (STDDS) with spatial and channel attention downsamples the shallower feature map without discarding detail; a small network then predicts an offset field and sampling weights from the concatenation of this downsampled shallow map and the deeper map; deformable convolution re-samples the deeper map so its target positions and edges match the shallow map. Since SPAM is applied repeatedly from the finest level to the coarsest, each deeper level inherits alignment back to the shallowest level, which is the mechanism the paper calls global alignment. After the top-down path begins, the Fine-grained Semantic Alignment Module (SEAM) computes channel-level and pixel-level global masks, root-combines them, and applies the product to the deep feature before adding it to the shallow feature, suppressing aliasing while preserving diversity. The Group Aggregation Lateral Connection Module (GALM) groups channels by shared semantics and learns aggregation weights, replacing channel-compressing 1x1 lateral connections.

What would settle it

Measure, per pyramid level, the center-point offset and shape distortion of the same objects in features produced by FPN and by BAFPN relative to the original image. If the bottom-up SPAM cascade does not reduce those errors at every deeper level, the claimed global-alignment mechanism is not what produces the AP75 gain. A second check is an ablation that replaces the bottom-up cascade with simple pairwise adjacent-level alignment; if AP75 stays the same, the 'global' part of the claim is not load-bearing.

Watch

Extended reading notes

Core claim

The central claim is that misalignment of deep features with the original image—positional shifts and shape distortion accumulated through many convolutions—limits high-IoU localization in rotated-object detection, and that a feature pyramid can repair this by aligning adjacent levels bottom-up before any top-down fusion. BAFPN implements this claim: the Spatial Feature Alignment Module (SPAM), chained from the finest backbone feature to the coarsest, resamples each deeper map to its shallower reference using deformable convolution driven by an offset field and sampling weights, so every level ends up aligned back to the lowest-level features. On DOTAv1.5, replacing FPN in Oriented R-CNN with BAFPN raises AP75 from 40.22% to 41.90%, AP50 from 66.67% to 68.12%, and mAP from 39.42% to 40.84%. The paper reports consistent AP75 gains on Rotated Faster R-CNN, Rotated RetinaNet, Rotated FCOS, Gliding Vertex, and RoI Transformer, with the largest gains for anchor-free detectors, and interprets this as evidence that correcting global feature alignment improves localization accuracy rather than merely adding parameters.

Load-bearing premise

The argument assumes the shallowest backbone feature map is already correctly aligned with the original image, because the entire bottom-up cascade uses that layer as its reference; if the finest feature is itself shifted or distorted, the chain cannot restore true global alignment.

Editorial extensions

If this is right

  • If BAFPN's alignment account is right, high-IoU metrics such as AP75 are the place to look for gains, and the reported +1.68 point AP75 over the Oriented R-CNN baseline is the direct signature of improved localization rather than improved classification.
  • The paper's account implies anchor-free detectors are the most sensitive to feature misalignment, since their center-ness and point-based predictions depend on where fused features land; Rotated FCOS gains about 3 points in AP when FPN is replaced by BAFPN.
  • Deepening the backbone should make the alignment problem worse, and the paper's ResNet101 result—roughly 1 point mAP and AP75 gain, over 2 points AP50 gain—fits that prediction.
  • Because BAFPN adds only about 3 MB over the PAFPN-based version and still beats NAS-FPN and PAFPN on AP75, the localization gain is attributed to the alignment mechanism rather than to model size.

Reading between the lines

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

  • Inference: if global alignment is the operative mechanism, the advantage should grow at even stricter overlap thresholds, so evaluating BAFPN at AP90 or with a localization-only metric would be a sharper test than the reported AP75.
  • Inference: the bottom-up cascade treats the shallowest feature map as the trusted reference; a test that shifts or distorts only that finest level would reveal how much residual error the chain can correct.
  • Inference: the same design could transfer to generic small-object detection, but the expected gain depends on how much geometric drift a backbone introduces; comparing BAFPN's AP75 delta across datasets with different backbone strides would separate the alignment effect from the fusion effect.
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 / 6 minor

Summary. The paper proposes BAFPN, a modified feature pyramid network for oriented object detection in remote sensing images. BAFPN adds a bottom-up spatial-alignment path in which a Spatial Feature Alignment Module (SPAM) aligns each adjacent feature map using Spatial-to-Depth Down-Sampling (STDDS) and deformable convolution, followed by a top-down Fine-Grained Semantic Alignment Module (SEAM) that generates channel-and-pixel-level masks, plus a Grouped Aggregation Lateral Connection Module (GALM) that replaces the standard 1x1 lateral convolution. On DOTAv1.5, the authors report that replacing FPN in Oriented R-CNN improves AP75 from 40.22 to 41.90, AP50 from 66.74 to 68.12, and mAP from 39.42 to 40.84, with additional gains on five other rotated detectors. The paper includes a module-level ablation, qualitative feature visualizations, and a comparison with NAS-FPN and PAFPN.

Significance. If the results are reproducible and the interpretation is accurate, BAFPN is a useful, low-cost improvement to FPN-style necks for oriented detection: the gains are consistent across six detectors and two backbones, the ablations support a role for each of the three modules, and the model-size overhead relative to PAFPN is small (about 3 MB). The paper also gives a clear explanation of how its components interact. The main weaknesses are the absence of quantitative evidence for the central 'global alignment to the original image' claim, small effect sizes without error bars, numerical inconsistencies between the abstract, text, and tables, and the lack of code or multi-seed runs. These issues prevent the paper from being accepted in its current form.

major comments (3)
  1. [III-B (and abstract, Section I)] The paper's central claim of 'global-scale' alignment rests on the assertion that aligning each feature level to the immediately shallower level, cascaded bottom-up, aligns all levels to the lowest-layer feature (P2). This is stated explicitly in Section III-B: 'SPAM only align two adjacent feature maps, but in BAFPN, the alignment process is performed from bottom to top, which allows the targets in each feature layer to adjust their spatial positions based on the features of the lowest layer.' The load-bearing premise is that P2 is itself aligned to the input image. The manuscript provides no quantitative evidence for this: Figures 7, 8, and 10 are selected qualitative visualizations, and there is no measurement of alignment error (e.g., centroid offset or mask IoU between feature activations and ground-truth object regions) at any pyramid level. If P2 carries a residual shift or distortion, the cascade propagates that bias and does not restore alignment to the original image. I request either (i) a direct quantitative alignment-error comparison between FPN and BAFPN at each level relative to ground-truth boxes or masks, or (ii) a revised formulation that claims only adjacent-level alignment and drops the 'global to original image' wording.
  2. [IV-C, Tables 1-2, abstract] The headline numbers are inconsistent. The abstract reports improvements of +1.68 AP75, +1.45 AP50, and +1.34 mAP over the Oriented R-CNN baseline. Table 2 lists baseline (ORC, ResNet50) as 40.22 AP75, 66.74 AP50, 39.42 mAP and ORC* as 41.90, 68.12, 40.84, giving differences of +1.68, +1.38, and +1.42. Section IV-C states AP50 increased from 66.67 to 68.12 and mAP from 39.6 to 40.84. These discrepancies must be reconciled because they directly affect the paper's quantitative claim.
  3. [IV-B, IV-C, Tables 1-3] All comparisons are single-run, single-dataset (DOTAv1.5), with no error bars, no multi-seed results, and no code release. Many reported gains are small: for example, Table 2 shows RFR* mAP 36.73 vs 35.98 (+0.75) and ORC* AP75 +1.68; Table 3 contains differences of 0.1-0.2 AP. Without variance estimates, it is not possible to assess whether these differences are statistically meaningful. The authors should report mean ± standard deviation over at least three seeds, or clearly qualify the results as single-run observations. This is important because the central claim is an improvement over existing FPN variants.
minor comments (6)
  1. [III-A] In the GALM equations, C_i^j = ψ1(Ĉ_i^j) uses C_i^j for both the input group and the output with different dimensions; please rename one of the variables to avoid confusion.
  2. [III-B] There are typos: 'Spatial-to-Deepth' should be 'Spatial-to-Depth', and Figure 3's caption reads 'Spacial Features Align Module' instead of 'Spatial Feature Alignment Module'.
  3. [III-C] The range for the saliency weakening factor is written as 'k(1 ≤ k ≤ 0)', which is impossible; it should be '0 ≤ k ≤ 1'.
  4. [IV-D, Table 3] The symbol 〇 in the SPAM column is explained in the caption, but the text refers to 'OEAM replaced with nearest-neighbor downsampling'; the notation should be made consistent (the intended module is SEAM or STDDS as appropriate).
  5. [IV-C, Figure 6] The paper compares model size versus AP75 but does not report inference speed or FLOPs; a full efficiency comparison would be more informative for a module intended to be lightweight.
  6. [Figures 7-8] The feature visualizations are selected examples; a quantitative overlay, such as centroid error or activation-IoU against ground-truth boxes, would strengthen the claim that BAFPN reduces position shift and shape distortion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's performance claims are empirical measurements on a public benchmark, and each module is defined by its own equations without any fitted quantity being relabeled as a prediction.

full rationale

The central claim of the paper is that BAFPN improves AP75, AP50, and mAP on DOTAv1.5 relative to an Oriented R-CNN baseline. This is an experimentally measured result on an external benchmark, not a quantity derived from the model's own fitted parameters. The three proposed modules (SPAM, SEAM, GALM) are each defined by explicit equations (Section III) and are trained end-to-end; no parameter is fit to a subset of data and then used to predict a closely related quantity. The paper contains no self-citations: all references are to prior external work (DCN, CBAM, SE-Net, FPN variants, detectors), and no load-bearing argument relies on the authors' own prior results or on an imported uniqueness theorem. The global-alignment claim in Section III-B is a qualitative architectural rationale supported by ablations and feature visualizations; even if one disputes whether the cascade truly restores global alignment or whether the shallowest layer is an undistorted reference, that is a correctness or verification concern, not a circularity. The ablation study in Table III isolates each module's contribution against a baseline, which is the standard empirical methodology and does not reduce to the paper's inputs by construction. No step in the derivation chain equates an output with an input by definition, and no fitted input is renamed as a prediction. Therefore the paper shows no significant circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 4 invented entities

The four modules are new architectural entities; the paper relies on two domain assumptions about feature misalignment and semantic aliasing. No physical constants or exotic entities are introduced.

free parameters (3)
  • Number of semantic groups g in GALM = not reported
    GALM divides channels into g groups; the paper does not state how g is chosen, and performance likely depends on it.
  • SEAM pixel-branch kernel size and group count = 7x7, C_out groups
    Kernel size and group count are chosen by hand with no sensitivity analysis.
  • STDDS output channel count = matched to deep feature
    The 1x1 convolution after space-to-depth compresses to the deep feature channel count, a design choice inherited from FPN.
assumptions (4)
  • domain assumption Deep backbone features accumulate positional shifts and shape distortions relative to the input image.
    This motivates the SPAM design and is asserted in Sections I and III-B, but it is not directly measured.
  • domain assumption Aligning adjacent pyramid levels bottom-up yields global alignment to the original image.
    The key premise, stated in Section III-B, that anchoring to the lowest layer fixes global misalignment.
  • domain assumption Semantic gaps between adjacent feature levels cause aliasing that hurts localization, and masking can reduce this without losing diversity.
    This motivates SEAM in Section III-C.
  • domain assumption AP50, AP75, and mAP on DOTAv1.5 are the right proxies for localization accuracy.
    The evaluation treats AP75 as the strict localization metric; no other localization measure is reported.
invented entities (4)
  • Spatial Feature Alignment Module (SPAM) independent evidence
    purpose: Aligns deeper feature maps to shallower ones via deformable convolution guided by STDDS downsampling.
    A new module whose effect is tested in ablation and main experiments on DOTAv1.5.
  • Spatial-to-Depth Down-Sampling (STDDS) independent evidence
    purpose: Downsamples shallow features while preserving spatial detail for alignment guidance.
    Ablation row replacing STDDS with a 3x3 convolution supports its contribution.
  • Fine-Grained Semantic Alignment Module (SEAM) independent evidence
    purpose: Generates global channel-pixel masks to reduce the semantic gap in top-down fusion.
    Ablation shows AP improvements attributed to SEAM.
  • Grouped Aggregation Lateral Connection Module (GALM) independent evidence
    purpose: Replaces 1x1 lateral convolution with grouped semantic aggregation to retain channel information.
    Ablation shows an AP75 gain of 0.78 points.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BAFPN: Bi directional alignment of features to improve localization accuracy." pith.science (2026). https://pith.science/paper/6I3DSTDN

@misc{pith2026241201859,
  author       = {Pith},
  title        = {Pith review of: BAFPN: Bi directional alignment of features to improve localization accuracy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6I3DSTDN}},
  note         = {Machine review of arXiv:2412.01859}
}
read the original abstract

Current state-of-the-art vision models often utilize feature pyramids to extract multi-scale information, with the Feature Pyramid Network (FPN) being one of the most widely used classic architectures. However, traditional FPNs and their variants (e.g., AUGFPN, PAFPN) fail to fully address spatial misalignment on a global scale, leading to suboptimal performance in high-precision localization of objects. In this paper, we propose a novel Bidirectional Alignment Feature Pyramid Network (BAFPN), which aligns misaligned features globally through a Spatial Feature Alignment Module (SPAM) during the bottom-up information propagation phase. Subsequently, it further mitigates aliasing effects caused by cross-scale feature fusion via a fine-grained Semantic Alignment Module (SEAM) in the top-down phase. On the DOTAv1.5 dataset, BAFPN improves the baseline model's AP75, AP50, and mAP by 1.68%, 1.45%, and 1.34%, respectively. Additionally, BAFPN demonstrates significant performance gains when applied to various other advanced detectors.

Figures

Figures reproduced from arXiv: 2412.01859 by the authors.

Figure 1
Figure 1. The overall structure of BAFPN and other FPNs II. RELATED WORK A. Rotated Object Detection Oriented object detection aims to recognize and locate objects rotated in arbitrary directions, which is particularly suitable for object detection in aerial and remote sensing images. Traditional object detection methods typically assume that objects are axis-aligned rectangles, which are not ideal for detecting directional o… view at source ↗
Figure 2
Figure 2. Group Aggregation Lateral Model In fact, the features output by the backbone network often contain many channels with similar semantics. If channels with similar semantics are grouped together first and then weighted and fused with channels of different semantics, the fusion efficiency can be significantly improved, and richer feature representations can be obtained. To this end, we propose a Grouped Aggregation Lat… view at source ↗
Figure 3
Figure 3. Spacial Features Align Module B. Spatial Feature Alignment at Global Scale The structure of the Spatial Feature Alignment Module (SPAM) is shown in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Spacial to Depth Down Sample These sub-sampling operations extract interleaved pixel blocks from the spatial dimensions of the input feature map. The four sub-sampled results are concatenated along the channel dimension to transform spatial information into depth infor…
Figure 5
Figure 5. Figure 5: Semantic Align Module Specifically, SEAM consists of a global context-weighting branch and a pixel-weighting branch. The global context￾weighting branch compresses the feature map along the channel dimension into four sets of global feature vectors through max￾pooling …
Figure 6
Figure 6. Figure 6: Model size vs. AP75 comparison To verify BAFPN's advantage in localization performance over other advanced FPN variants, we replaced the FPN in Oriented RCNN with two advanced plug-and-play FPNs: PAFPN [38] and NAS-FPN, and compared them with the BAFPN detector. The NA…
Figure 8
Figure 8. Figure 8: Feature Shifting and Correction in FPN To visually demonstrate the advantages of BAFPN in feature alignment, we compared the feature visualization of FPN and BAFPN at the same level. As shown in Figures 7a and 7b, the features extracted from FPN exhibit significant dis…
Figure 7
Figure 7. Figure 7: Feature Distortion and Repair in FPN To verify the necessity of global feature alignment in improving the detector's localization accuracy, we retained only the SPAM in BAFPN and replaced STDDS with a 3×3 convolution with a stride of 2. In this case, BAFPN becomes equi…
Figure 10
Figure 10. Figure 10: Visualization of Space-to-Depth Downsampled Features V. CONCLUSION This paper provides a detailed analysis of the limitations of FPN and its various variants in remote sensing image object detection, and offers a comprehensive improvement by proposing a novel BAFPN. B…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 35 canonical work pages

  1. [1]

    Feature pyramid networks for object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Lin T Y, Dollá r P, Girshick R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 2117-2125

    work page 2017

  2. [2]

    Info-FPN: An Informative Feature Pyramid Network for object detection in remote sensing images[J]

    Chen S, Zhao J, Zhou Y, et al. Info-FPN: An Informative Feature Pyramid Network for object detection in remote sensing images[J]. Expert Systems with Applications, 2023, 214: 119132

    work page 2023

  3. [3]

    CE -FPN: enhancing channel information for object detection[J]

    Luo Y, Cao X, Zhang J, et al. CE -FPN: enhancing channel information for object detection[J]. Multimedia Tools and Applications, 2022, 81(21): 30685-30704

    work page 2022

  4. [4]

    Fapn: Feature -aligned pyramid network for dense image prediction[C]//Proceedings of the IEEE/CVF international conference on computer vision

    Huang S, Lu Z, Cheng R, et al. Fapn: Feature -aligned pyramid network for dense image prediction[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2021: 864-873

    work page 2021

  5. [5]

    Augfpn: Improving multi -scale feature learning for object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Guo C, Fan B, Zhang Q, et al. Augfpn: Improving multi -scale feature learning for object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 12595 - 12604

    work page 2020

  6. [6]

    A2-FPN: Attention aggregation based feature pyramid network for instance segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Hu M, Li Y, Fang L, et al. A2-FPN: Attention aggregation based feature pyramid network for instance segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021: 15343-15352

    work page 2021

  7. [7]

    Feature alignment FPN for oriented object detection in remote sensing images[J]

    Li Z, Li E, Xu T, et al. Feature alignment FPN for oriented object detection in remote sensing images[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20: 1-5

    work page 2023

  8. [8]

    Arbitrary -oriented scene text detection via rotation proposals[J]

    Ma J, Shao W, Ye H, et al. Arbitrary -oriented scene text detection via rotation proposals[J]. IEEE transactions on multimedia, 2018, 20(11): 3111-3122

    work page 2018

Show all 40 references
  1. [9]

    H2rbox: Horizontal box annotation is all you need for oriented object detection[J]

    Yang X, Zhang G, Li W, et al. H2rbox: Horizontal box annotation is all you need for oriented object detection[J]. arXiv preprint arXiv:2210.06742, 2022

    2022 arXiv

  2. [10]

    Building a bridge of bounding box regression between oriented and horizontal object detection in remote sensing images[J]

    Qian X, Wu B, Cheng G, et al. Building a bridge of bounding box regression between oriented and horizontal object detection in remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1-9

    2023

  3. [11]

    Arbitrary-oriented object detection with circular smooth label[C]//Computer Vision –ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part VIII 16

    Yang X, Yan J. Arbitrary-oriented object detection with circular smooth label[C]//Computer Vision –ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part VIII 16. Springer International Publishing, 2020: 677-694

    2020

  4. [12]

    Task interleaving and orientation estimation for high-precision oriented object detection in aerial images[J]

    Ming Q, Miao L, Zhou Z, et al. Task interleaving and orientation estimation for high-precision oriented object detection in aerial images[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 196: 241- 255

    2023

  5. [13]

    Rethinking rotated object detection with gaussian wasserstein distance loss[C]//International conference on machine learning

    Yang X, Yan J, Ming Q, et al. Rethinking rotated object detection with gaussian wasserstein distance loss[C]//International conference on machine learning. PMLR, 2021: 11830-11841

    2021

  6. [14]

    Learning modulated loss for rotated object detection[C]//Proceedings of the AAAI conference on artificial intelligence

    Qian W, Yang X, Peng S, et al. Learning modulated loss for rotated object detection[C]//Proceedings of the AAAI conference on artificial intelligence. 2021, 35(3): 2458-2466

    2021

  7. [15]

    Kim B, Lee J, Lee S, et al. TricubeNet: 2D kernel -based object representation for weakly -occluded oriented object detection[C]//Proceedings of the IEEE/CVF winter conference on applications of computer vision. 2022: 167-176

    2022

  8. [16]

    Adaptive rotated convolution for rotated object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision

    Pu Y, Wang Y, Xia Z, et al. Adaptive rotated convolution for rotated object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023: 6589-6600

    2023

  9. [17]

    Large selective kernel network for remote sensing object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision

    Li Y, Hou Q, Zheng Z, et al. Large selective kernel network for remote sensing object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023: 16794-16805

    2023

  10. [19]

    A2-FPN for semantic segmentation of fine- resolution remotely sensed images[J]

    Li R, Wang L, Zhang C, et al. A2-FPN for semantic segmentation of fine- resolution remotely sensed images[J]. International journal of remote sensing, 2022, 43(3): 1131-1155

    2022

  11. [20]

    Cross -layer feature pyramid network for salient object detection[J]

    Li Z, Lang C, Liew J H, et al. Cross -layer feature pyramid network for salient object detection[J]. IEEE Transactions on Image Processing, 2021, 30: 4587-4598

    2021

  12. [21]

    Lr -fpn: Enhancing remote sensing object detection with location refined feature pyramid network[J]

    Li H, Zhang R, Pan Y, et al. Lr -fpn: Enhancing remote sensing object detection with location refined feature pyramid network[J]. arXiv preprint arXiv:2404.01614, 2024

    2024 arXiv

  13. [22]

    ssFPN: Scale sequence (S 2) feature-based feature pyramid network for object detection[J]

    Park H J, Kang J W, Kim B G. ssFPN: Scale sequence (S 2) feature-based feature pyramid network for object detection[J]. Sensors, 2023, 23(9): 4432

    2023

  14. [23]

    Detectors: Detecting objects with recursive feature pyramid and switchable atrous convolution[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Qiao S, Chen L C, Yuille A. Detectors: Detecting objects with recursive feature pyramid and switchable atrous convolution[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021: 10213-10224

    2021

  15. [24]

    Nas -fpn: Learning scalable feature pyramid architecture for object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Ghiasi G, Lin T Y, Le Q V. Nas -fpn: Learning scalable feature pyramid architecture for object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 7036-7045

    2019

  16. [25]

    Auto-fpn: Automatic network architecture adaptation for object detection beyond classification[C]//Proceedings of the IEEE/CVF international conference on computer vision

    Xu H, Yao L, Zhang W, et al. Auto-fpn: Automatic network architecture adaptation for object detection beyond classification[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 6649- 6658

    2019

  17. [26]

    Effective fusion factor in FPN for tiny object detection[C]//Proceedings of the IEEE/CVF winter conference on applications of computer vision

    Gong Y, Yu X, Ding Y, et al. Effective fusion factor in FPN for tiny object detection[C]//Proceedings of the IEEE/CVF winter conference on applications of computer vision. 2021: 1160-1168

    2021

  18. [27]

    Retro-fpn: Retrospective feature pyramid network for point cloud semantic segmentation[C]//Proceedings of the IEEE/CVF international conference on computer vision

    Xiang P, Wen X, Liu Y S, et al. Retro-fpn: Retrospective feature pyramid network for point cloud semantic segmentation[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2023: 17826 - 17838

    2023

  19. [28]

    Deformable convolutional networks[C]//Proceedings of the IEEE international conference on computer vision

    Dai J, Qi H, Xiong Y, et al. Deformable convolutional networks[C]//Proceedings of the IEEE international conference on computer vision. 2017: 764-773

    2017

  20. [29]

    Cbam: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV)

    Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 3-19

    2018

  21. [30]

    Squeeze -and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Hu J, Shen L, Sun G. Squeeze -and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141

    2018

  22. [31]

    Faster r -cnn: Towards real -time object detection with region proposal networks[J]

    Ren S. Faster r -cnn: Towards real -time object detection with region proposal networks[J]. arXiv preprint arXiv:1506.01497, 2015

    2015 arXiv

  23. [32]

    Learning RoI transformer for oriented object detection in aerial images[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Ding J, Xue N, Long Y, et al. Learning RoI transformer for oriented object detection in aerial images[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 2849-2858

    2019

  24. [33]

    Gliding vertex on the horizontal bounding box for multi -oriented object detection[J]

    Xu Y, Fu M, Wang Q, et al. Gliding vertex on the horizontal bounding box for multi -oriented object detection[J]. IEEE transactions on pattern analysis and machine intelligence, 2020, 43(4): 1452-1459

    2020

  25. [34]

    Focal Loss for Dense Object Detection[J]

    Lin T. Focal Loss for Dense Object Detection[J]. arXiv preprint arXiv:1708.02002, 2017

    2017 arXiv

  26. [35]

    FCOS: Fully convolutional one -stage object detection

    Tian Z, Shen C, Chen H, et al. FCOS: Fully convolutional one -stage object detection. arXiv 2019[J]. arXiv preprint arXiv:1904.01355, 2019

    2019 arXiv

  27. [36]

    Oriented R -CNN for object detection[C]//Proceedings of the IEEE/CVF international conference on computer vision

    Xie X, Cheng G, Wang J, et al. Oriented R -CNN for object detection[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2021: 3520-3529

    2021

  28. [37]

    DOTA: A large -scale dataset for object detection in aerial images[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Xia G S, Bai X, Ding J, et al. DOTA: A large -scale dataset for object detection in aerial images[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 3974-3983

    2018

  29. [38]

    Path aggregation network for instance segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    Liu S, Qi L, Qin H, et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 8759-8768

    2018

  30. [39]

    Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition

    He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778

    2016

  31. [40]

    AFPN: Asymptotic feature pyramid network for object detection[C]//2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

    Yang G, Lei J, Zhu Z, et al. AFPN: Asymptotic feature pyramid network for object detection[C]//2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC). IEEE, 2023: 2184-2189

    2023

  32. [41]

    Improving language models by retrieving from trillions of tokens[C]//International conference on machine learning

    Borgeaud S, Mensch A, Hoffmann J, et al. Improving language models by retrieving from trillions of tokens[C]//International conference on machine learning. PMLR, 2022: 2206-2240. Jiakun Li: Bachelor of Engineering , graduated from Shandong University of Science and Technology,...

    2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.