Corner2Net: Detecting Objects as Cascade Corners

Chenglong Liu; Haorao Wei; Jintao Liu; Jinze Yang; Liangyu Xu; Lu Fang; Yuchen Guo

REVIEW 3 major objections 4 minor 33 references

Corner2Net: Detecting Objects as Cascade Corners

T0 review · 3 major / 4 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 Corner2Net models an object as two cascade corners — a class-agnostic top-left corner found first, then an instance-specific bottom-right corner located inside the region it defines — and reports 47.8% AP on the COCO test-dev, ahead of…

desk verdict A genuine new cascade-corner idea with honest ablations, but the 'large margin' SOTA claim is overstated once you notice the backbones differ and the gap to CornerAffinity is ~0.5 AP. read the letter →

arxiv 2411.15772 v1 pith:VLPAQR2D submitted 2024-11-24 cs.CV

Chenglong Liu , Jintao Liu , Haorao Wei , Jinze Yang , Liangyu Xu , Yuchen Guo , Lu Fang This is my paper · ORCID

classification cs.CV

keywords objectdetectioncorner-basedcascadecornerskeypointheatmapsRoIfeatureclassificationCOCOcornermatchingtwo-stagedetector

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

Corner2Net is a proposal to rehabilitate corner-based object detection by changing the order in which corners are found. Instead of predicting top-left and bottom-right corners independently and then matching them heuristically, the detector first finds all class-agnostic top-left corners, uses each to define a square search region scaled by the object's estimated size, and then locates that instance's bottom-right corner inside that region. Object category is assigned separately from rich region-of-interest (RoI) features, so the two corners no longer need to carry class semantics. On the COCO test-dev set the paper reports 47.8% average precision under multi-scale testing, above all previously published corner-based detectors, while running about 2.1 times faster than the CornerNet baseline it builds on. The significance, if the results hold, is that the corner-based paradigm can use standard backbones and avoid fragile matching without giving up accuracy.

What carries the argument

The load-bearing mechanism is the cascade corner association via an enlarged square RoI: a top-left corner and a regressed object scale define a proposal box, and enlarging that scale by $\eta=0.25$ creates the region inside which the second stage must find that instance's bottom-right corner. This RoI is what links one corner to the other, so the bottom-right corner is no longer searched over the whole image but only inside a region anchored by its partner. Supporting components are the class-agnostic focal loss with distance-aware penalty, bottom-right corner pooling plus deconvolution for the instance-level heatmap, GIoU loss for the scale regression, Smooth L1 losses for the two offsets, and a RoI-feature classification head.

What would settle it

The paper itself provides the sharpest test: setting the enlarge factor to 0 drops COCO AP from 45.7 to 32.5 on val2017, showing how much the result depends on the enlarged RoI containing the bottom-right corner. A more direct check would be to count, on a labeled set, how often the true bottom-right corner lies outside the predicted RoI and to measure top-left corner recall independently; if either is often wrong, objects are lost before the second stage begins.

Watch

Extended reading notes

Core claim

The central claim is that the two corners of a bounding box should be treated as a cascade, not as a pair of parallel predictions. The first stage outputs a single-channel heatmap of class-agnostic top-left corners on FPN levels, together with position offsets and a per-instance scale; from that it decodes a proposal box with square side length equal to the predicted scale, enlarged by a factor of 0.25 to form an RoI that should contain the true bottom-right corner. The second stage applies RoIAlign and bottom-right corner pooling to produce a 28 by 28 instance-specific heatmap, picks the peak to locate the bottom-right corner, and refines it with an offset. Classification uses the un-enlarged RoI features through a lightweight head, with the final score combining localization confidence and class confidence. The paper claims this design makes corner matching unnecessary, decouples localization from classification, and yields 47.8% AP on COCO test-dev with a ResNeXt-101-DCN backbone, 46.0% with Swin-T, and large gains over CornerNet on crowded and aerial benchmarks.

Load-bearing premise

The whole pipeline depends on the first stage finding every top-left corner and on the estimated object scale, enlarged by 25%, producing a square region that contains the true bottom-right corner; if either fails, the second stage has no way to recover the object.

Editorial extensions

If this is right

Corner-based detectors no longer need a corner-matching post-process or an instance embedding; a detected box is just a top-left peak plus the bottom-right peak found inside its RoI.
Because the first-stage corners are class-agnostic, the detector can run on a single-channel heatmap and connect to standard FPN backbones, so training takes 24 epochs instead of the 200 to 320 epochs required by hourglass-based corner detectors.
Decoupling classification from the corner heatmaps should improve robustness when similar-looking objects co-occur, which is exactly where the paper reports its largest gains on COCO, CityPersons, and UCAS-AOD.
The reported AP80 of 44.6% and AP90 of 22.4% suggest the cascade produces tightly fitting boxes, consistent with the idea that corners are a precise box representation.

Reading between the lines

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

A testable extension of the cascade idea is to replace the single square scale with separate width and height regressions; the square RoI's dependence on one scale is the main internal bottleneck identified by the enlarge-factor ablation, and anisotropic RoIs could recover extreme-aspect-ratio objects.
The class-agnostic first stage could potentially be reused across datasets or even across tasks such as oriented boxes or panoptic segmentation, since only the second-stage head would need retraining.
If the first stage misses a top-left corner, the object is unrecoverable; one repair the paper does not explore is a second pass or a top-k set of top-left candidates per scale, which could be evaluated directly by measuring top-left corner recall.

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.

Desk Editor's Note

read the letter

The one thing to know: this is not a recycled corner detector. Corner2Net's cascade pipeline—class-agnostic top-left corner first, then an instance-specific bottom-right corner searched inside the resulting RoI—is genuinely new relative to CornerNet, CenterNet, CentripetalNet, and CornerAffinity, and it eliminates the fragile corner-matching post-process. That's a real conceptual step, and the paper does it cleanly. The decoupling of localization from classification via RoI features is sensible, and the design lets them swap in common backbones like ResNet and Swin-Transformer, which prior parallel corner detectors couldn't do without the Hourglass. The ablation study is honest: Table 5 shows the enlarge factor eta matters a lot (AP drops from 45.7 to 32.5 at eta=0), and Table 4 shows insensitivity to top-k, which is useful information.

Where the paper oversells: the 'large margin' claim. On COCO test-dev, Corner2Net gets 46.8 AP with ResNeXt-101-DCN, while CornerAffinity gets 46.3 with Hourglass-104. That's 0.5 AP, and the backbones differ, so the comparison is not controlled. The speed comparison has the same confound—of course a ResNeXt model with FPN runs faster than Hourglass-104. Also, the CityPersons filtering protocol is described in one sentence and not reproducible, there are no error bars or multiple runs reported, and no code is released. The stress-test concern about serial error accumulation is real: if the first stage misses a top-left corner or underestimates scale beyond the eta slack, the second stage can't recover. That's a structural property of the method, not a flaw in the math, but the paper never quantifies top-left recall or scale-error distribution, so we don't know how often this bottleneck bites.

Overall, the method is plausible and the design is well-motivated. The paper deserves a serious referee and could become a useful baseline for corner-based detection. The needed revision is mostly about evaluation transparency: report variance, document CityPersons filtering, release code, and soften the 'large margin' language to something the evidence actually supports. I'd take a look at a revised version, and I'd probably cite it as a baseline in future work.

Referee Report

3 major / 4 minor

Summary. The paper proposes Corner2Net, a two-stage cascade corner-based object detector. In the first stage, class-agnostic top-left corners are detected on FPN heatmaps along with a square RoI defined by a regressed object scale; in the second stage, the instance-specific bottom-right corner is localized within the enlarged RoI and the object category is predicted from RoI features. The method is evaluated on MS-COCO, CityPersons, and UCAS-AOD. The authors report 46.8/47.8 AP on COCO test-dev with ResNeXt-101-DCN under single/multi-scale testing and claim to surpass all existing corner-based detectors in accuracy and speed.

Significance. The cascade-corner formulation is a genuinely different way to sidestep heuristic corner matching, and the paper demonstrates that it can be trained with standard backbones in far fewer epochs than Hourglass-based corner detectors. The ablation on the enlarge factor (Table 5) is honest and informative, and the high-AP80 results are a useful data point. The speed advantage (8.0 FPS versus 3.7-3.9 FPS for Hourglass-based corner detectors) is meaningful. However, the central 'large margin' claim is overstated: the single-scale AP gap over the best prior corner-based detector (CornerAffinity) is only 0.5 AP, and the cascade's dependence on first-stage top-left recall and scale accuracy is not quantified, leaving the robustness of the method under-specified.

major comments (3)

[Section 1, Table 1] The claim that Corner2Net 'exceeds all existing corner-based detectors by a large margin' is not supported by the data in Table 1 for single-scale testing: Corner2Net achieves 46.8 AP on COCO test-dev versus 46.3 AP for CornerAffinity, a 0.5 AP difference, under a different backbone (ResNeXt-101-DCN versus Hourglass-104) and much shorter training (24 versus 320 epochs). Even the multi-scale result (47.8 versus 46.3) is a 1.5 AP gap rather than a large margin, and no error bars are reported for any configuration. Please either provide controlled comparisons (same backbone and schedule) and repeated-run variance, or temper the claim to 'comparable or slightly better accuracy with substantially higher speed and training efficiency.'
[Section 4.6, Table 5; Section 4.2] The cascade pipeline's accuracy is critically dependent on the first-stage top-left corner and scale estimate, as shown by the eta ablation: setting eta=0 drops AP from 45.7 to 32.5 on COCO val2017, and eta=0.10 still loses 0.7 AP relative to eta=0.25. The paper does not report the recall of the first-stage top-left detector at the top-128 truncation, the distribution of scale regression errors, or the fraction of instances whose true bottom-right corner falls outside the un-enlarged RoI. Without these diagnostics, the reader cannot separate first-stage error accumulation from the method's localization quality; given the small accuracy margin over prior corner detectors, this gap undermines the robustness of the state-of-the-art claim. Please add these analyses and discuss how often the second stage can recover from first-stage misses.
[Section 4.1, Table 3] The CityPersons evaluation protocol is not sufficiently described. The sentence 'We filter the vanilla dataset, and merge the main annotations of pedestrians and riders' does not specify which images or annotations are removed, whether the standard 'reasonable' subset and evaluation settings are used, or whether AP is computed on visible or full-body boxes. Because Table 3 compares against methods whose CityPersons numbers may have been obtained under different protocols, the reported improvements of 36.2% and 18.0% over CornerNet on AP50 are not verifiable. Please specify the exact filtering, annotation merging, evaluation metric, and test split, and preferably use the standard CityPersons evaluation protocol.

minor comments (4)

[Eq. (3)] In the GIoU loss formula, the arguments of LGIoU appear to be duplicated: it reads LGIoU[(zj, zj), (zf_j, zf_j)] instead of comparing the predicted scale zj with the ground truth zf_j. Please correct the notation.
[Abstract, Section 4.3, Table 2] The abstract highlights 'remarkable AP80/AP90 of 44.6%/22.4%', but Table 2 shows that the AP90 of Corner2Net (22.4%) is lower than that of CornerNet (23.4%). The main text acknowledges this; the abstract should be adjusted to avoid overstating high-IoU quality.
[Section 4.2] Please clarify whether the top-128 proposal keypoints are selected per FPN level or globally across all levels, and how multi-level top-left predictions are merged before RoI extraction.
[Table 1] Two different methods named 'CenterNet' are listed (reference [30] as center-based and reference [5] as parallel corner-based). The text and table should disambiguate these (e.g., 'CenterNet (Objects as Points)' and 'CenterNet (Keypoint Triplets)') to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Corner2Net's claims are benchmarked against external COCO results, and the cascade pipeline is a learned compositional architecture rather than a derivation that re-imports its target.

full rationale

The paper makes no first-principles derivation; its central claims are empirical AP and FPS numbers on COCO, CityPersons, and UCAS-AOD. The two-stage cascade decodes the bottom-right corner from a learned heatmap inside an RoI determined by the predicted top-left corner and scale (Eqs. 4, 5, and 8). This is a compositional network design, not a tautology: the training targets are external ground-truth annotations, and the inference outputs are produced by learned heads, not by substituting the target into the formula. The enlarge factor eta is tuned on COCO val (Table 5) and then applied at test time; tuning a hyperparameter on a validation split is standard practice, not a fitted input called a prediction. The paper's own Table 5 honestly shows that the pipeline degrades when eta = 0, which is a limitation of first-stage accuracy rather than a circular step. The self-citations ([13], [14], [25], [26]) are prior published baselines or general references and are not load-bearing: no uniqueness theorem or ansatz is imported via citation, and the comparison numbers for CornerAffinity come from an externally published detector. The state-of-the-art claim rests on external benchmark comparisons and is therefore self-contained with respect to the reported results.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard deep learning assumptions and a few hyperparameters tuned on validation data. No new theoretical postulates or invented entities are introduced; the contribution is architectural and empirical.

free parameters (4)

Enlarge factor eta = 0.25
Tuned on COCO val2017 (Table 5). It controls how much the square RoI is expanded to ensure the bottom-right corner falls inside. AP drops sharply when eta=0, showing the result depends critically on this choice.
Top-k cascade corner pairs = 128
Selected based on the ablation in Table 4; the method is not very sensitive to this value, but it affects speed and recall.
Focal loss hyperparameters alpha and beta = 2 and 4
Taken from CornerNet rather than refit in this paper; they are inherited design choices that influence heatmap training.
Heatmap side length m = 28
Fixed output resolution for the bottom-right heatmap; affects quantization error and precision, as acknowledged in the AP90 discussion.

assumptions (5)

domain assumption ImageNet-pretrained backbone features transfer to object detection.
The method initializes backbones with ImageNet weights (Section 4.2), assuming this pretraining provides useful visual features.
domain assumption FPN multi-scale feature assignment based on object size is beneficial.
Objects are assigned to FPN levels following FCOS-like strategies (Section 3.1), which assumes size-based level assignment is a valid inductive bias.
domain assumption Corner pooling encodes useful boundary evidence.
Corner pooling is adopted from CornerNet (Section 3.2), assuming it improves corner localization.
standard math Standard backpropagation and stochastic gradient optimization converge to good local minima.
The training pipeline uses standard SGD and loss backpropagation without formal convergence guarantees.
domain assumption Objects are represented as axis-aligned bounding boxes.
The datasets and evaluation protocol define objects as axis-aligned boxes, which is a modeling choice of the benchmarks.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Corner2Net: Detecting Objects as Cascade Corners." pith.science (2026). https://pith.science/paper/VLPAQR2D

@misc{pith2026241115772,
  author       = {Pith},
  title        = {Pith review of: Corner2Net: Detecting Objects as Cascade Corners},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VLPAQR2D}},
  note         = {Machine review of arXiv:2411.15772}
}

read the original abstract

The corner-based detection paradigm enjoys the potential to produce high-quality boxes. But the development is constrained by three factors: 1) Hard to match corners. Heuristic corner matching algorithms can lead to incorrect boxes, especially when similar-looking objects co-occur. 2) Poor instance context. Two separate corners preserve few instance semantics, so it is difficult to guarantee getting both two class-specific corners on the same heatmap channel. 3) Unfriendly backbone. The training cost of the hourglass network is high. Accordingly, we build a novel corner-based framework, named Corner2Net. To achieve the corner-matching-free manner, we devise the cascade corner pipeline which progressively predicts the associated corner pair in two steps instead of synchronously searching two independent corners via parallel heads. Corner2Net decouples corner localization and object classification. Both two corners are class-agnostic and the instance-specific bottom-right corner further simplifies its search space. Meanwhile, RoI features with rich semantics are extracted for classification. Popular backbones (e.g., ResNeXt) can be easily connected to Corner2Net. Experimental results on COCO show Corner2Net surpasses all existing corner-based detectors by a large margin in accuracy and speed.

Figures

Figures reproduced from arXiv: 2411.15772 by the authors.

**Figure 1.** Comparision of parallel corner detection and the proposed cascade corner detection. All existing corner-based methods fall into the parallel detection pipeline, which predicts two separate class-specific corners and relies on the corner-matching algorithm when decoding boxes. Corner2Net adopts the proposed cascade corner detection pipeline that decouples the corner localization and object classification. Corner2N… view at source ↗

**Figure 2.** The framework of the proposed Corner2Net. Corner2Net has two cascade stages to conquer the corner-based detection task. In the first stage, all class-agnostic top-left corners are located on the image-level heatmap and adjusted by offsets, and each RoI space is determined to establish an association between two cascade corners. Next, the RoI features are fed into the second stage, where the precise bottom-right corn… view at source ↗

**Figure 2.** Clearly, our solution devises two cascade corners (or an as [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Figures from the paper (2 more)

**Figure 3.** Figure 3: Qualitative detection results of CornerNet baseline and the proposed Corner2Net on MS-COCO val2017 set [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]

**Figure 4.** Figure 4: Visualization results on UCAS-AOD (left column) and CityPersons (right column) datasets. boxes when similar objects co-occur or are partially occluded. This indicates that our model enjoys strong robustness and practicality. 5 Conclusion In this paper, we deeply analy…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 26 canonical work pages

[1]

Cai and N

Z. Cai and N. Vasconcelos. Cascade r-cnn: Delving into high quality object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6154–6162, 2018

work page 2018
[2]

Carion, F

N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko. End-to-end object detection with transformers. In Com- puter Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part I 16, pages 213–229. Springer, 2020

work page 2020
[3]

Cordts, M

M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benen- son, U. Franke, S. Roth, and B. Schiele. The cityscapes dataset for semantic urban scene understanding. In Proceedings of the IEEE con- ference on computer vision and pattern recognition, pages 3213–3223, 2016

work page 2016
[4]

Z. Dong, G. Li, Y . Liao, F. Wang, P. Ren, and C. Qian. Centripetal- net: Pursuing high-quality keypoint pairs for object detection. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10519–10528, 2020

work page 2020
[5]

K. Duan, S. Bai, L. Xie, H. Qi, Q. Huang, and Q. Tian. Centernet: Keypoint triplets for object detection. In Proceedings of the IEEE/CVF international conference on computer vision, pages 6569–6578, 2019

work page 2019
[6]

Girshick

R. Girshick. Fast r-cnn. In Proceedings of the IEEE international con- ference on computer vision, pages 1440–1448, 2015

work page 2015
[7]

K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

2016
[8]

K. He, G. Gkioxari, P. Dollár, and R. Girshick. Mask r-cnn. In Proceed- ings of the IEEE international conference on computer vision , pages 2961–2969, 2017

work page 2017

Show all 33 references

[9]

Krizhevsky, I

A. Krizhevsky, I. Sutskever, and G. E. Hinton. Imagenet classification with deep convolutional neural networks.Communications of the ACM, 60(6):84–90, 2017

2017
[10]

Law and J

H. Law and J. Deng. Cornernet: Detecting objects as paired keypoints. In Proceedings of the European conference on computer vision (ECCV), pages 734–750, 2018

2018
[11]

T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, and C. L. Zitnick. Microsoft coco: Common objects in con- text. In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages 740–7...

2014
[12]

T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Dollár. Focal loss for dense object detection. In Proceedings of the IEEE international con- ference on computer vision, pages 2980–2988, 2017

2017
[13]

C. Liu, H. Yu, H. Wei, X. Sun, and K. Fu. S2cgnet: A robust aircraft detector based on the sword-shaped component geometry. IEEE Trans- actions on Geoscience and Remote Sensing, 61:1–15, 2023

2023
[14]

C. Liu, H. Wei, J. Yang, J. Liu, W. Li, Y . Guo, and L. Fang. Gigahuman- det: Exploring full-body detection on gigapixel-level images. In Pro- ceedings of the AAAI Conference on Artificial Intelligence , volume 38, pages 10092–10100, 2024

2024
[15]

Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo. Swin transformer: Hierarchical vision transformer using shifted win- dows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021

2021
[16]

X. Lu, B. Li, Y . Yue, Q. Li, and J. Yan. Grid r-cnn. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 7363–7372, 2019

2019
[17]

Newell, K

A. Newell, K. Yang, and J. Deng. Stacked hourglass networks for hu- man pose estimation. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Pro- ceedings, Part VIII 14, pages 483–499. Springer, 2016

2016
[18]

J. Pang, K. Chen, J. Shi, H. Feng, W. Ouyang, and D. Lin. Libra r- cnn: Towards balanced learning for object detection. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 821–830, 2019

2019
[19]

Redmon and A

J. Redmon and A. Farhadi. Yolov3: An incremental improvement.arXiv preprint arXiv:1804.02767, 2018

2018 arXiv
[20]

Redmon, S

J. Redmon, S. Divvala, R. Girshick, and A. Farhadi. You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 779–788, 2016

2016
[21]

S. Ren, K. He, R. Girshick, and J. Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. Advances in neural information processing systems, 28, 2015

2015
[22]

Rezatofighi, N

H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid, and S. Savarese. Generalized intersection over union: A metric and a loss for bounding box regression. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 658–666, 2019

2019
[23]

P. Sun, R. Zhang, Y . Jiang, T. Kong, C. Xu, W. Zhan, M. Tomizuka, L. Li, Z. Yuan, C. Wang, et al. Sparse r-cnn: End-to-end object detection with learnable proposals. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14454–14463, 2021

2021
[24]

Z. Tian, C. Shen, H. Chen, and T. He. Fcos: Fully convolutional one- stage object detection. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9627–9636, 2019

2019
[25]

H. Wei, P. Guo, Y . Zhu, C. Liu, and P. Wang. Humanliker: A human- like object detector to model the manual labeling process. Advances in Neural Information Processing Systems, 35:2294–2306, 2022

2022
[26]

H. Wei, C. Liu, P. Guo, Y . Zhu, J. Fu, B. Wang, and P. Wang. Corner affinity: A robust grouping algorithm to make corner-guided detector great again. In IJCAI, pages 1458–1464, 2022

2022
[27]

S. Xie, R. Girshick, P. Dollár, Z. Tu, and K. He. Aggregated residual transformations for deep neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 1492– 1500, 2017

2017
[28]

Z. Yang, S. Liu, H. Hu, L. Wang, and S. Lin. Reppoints: Point set representation for object detection. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9657–9666, 2019

2019
[29]

Zhang, H

H. Zhang, H. Chang, B. Ma, N. Wang, and X. Chen. Dynamic r-cnn: To- wards high quality object detection via dynamic training. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XV 16, pages 260–275. Springer, 2020

2020
[30]

X. Zhou, D. Wang, and P. Krähenbühl. Objects as points. arXiv preprint arXiv:1904.07850, 2019

1904 arXiv
[31]

X. Zhou, V . Koltun, and P. Krähenbühl. Probabilistic two-stage detec- tion. arXiv preprint arXiv:2103.07461, 2021

2021 arXiv
[32]

C. Zhu, F. Chen, Z. Shen, and M. Savvides. Soft anchor-point object de- tection. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IX 16, pages 91–

2020
[33]

H. Zhu, X. Chen, W. Dai, K. Fu, Q. Ye, and J. Jiao. Orientation ro- bust object detection in aerial images using deep convolutional neural network. In 2015 IEEE International Conference on Image Processing (ICIP), pages 3735–3739. IEEE, 2015

2015

Pith tools

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

[1] [1]

Cai and N

Z. Cai and N. Vasconcelos. Cascade r-cnn: Delving into high quality object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6154–6162, 2018

work page 2018

[2] [2]

Carion, F

N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko. End-to-end object detection with transformers. In Com- puter Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part I 16, pages 213–229. Springer, 2020

work page 2020

[3] [3]

Cordts, M

M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benen- son, U. Franke, S. Roth, and B. Schiele. The cityscapes dataset for semantic urban scene understanding. In Proceedings of the IEEE con- ference on computer vision and pattern recognition, pages 3213–3223, 2016

work page 2016

[4] [4]

Z. Dong, G. Li, Y . Liao, F. Wang, P. Ren, and C. Qian. Centripetal- net: Pursuing high-quality keypoint pairs for object detection. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10519–10528, 2020

work page 2020

[5] [5]

K. Duan, S. Bai, L. Xie, H. Qi, Q. Huang, and Q. Tian. Centernet: Keypoint triplets for object detection. In Proceedings of the IEEE/CVF international conference on computer vision, pages 6569–6578, 2019

work page 2019

[6] [6]

Girshick

R. Girshick. Fast r-cnn. In Proceedings of the IEEE international con- ference on computer vision, pages 1440–1448, 2015

work page 2015

[7] [7]

K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

2016

[8] [8]

K. He, G. Gkioxari, P. Dollár, and R. Girshick. Mask r-cnn. In Proceed- ings of the IEEE international conference on computer vision , pages 2961–2969, 2017

work page 2017