Pith. sign in

REVIEW 3 major objections 5 minor 65 references

Long-Tailed Object Detection Pre-training: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction

T0 review · 3 major / 5 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 This paper proposes 2DRCL, a detection-specific pre-training method that combines holistic-local contrastive learning, dynamic rebalancing, and dual reconstruction to substantially improve rare-class accuracy in long-tailed object…

desk verdict Useful combination of known ideas, but the LVIS comparison is confounded by an unspecified pre-training protocol and an undefined Ldet; needs major revision, not desk reject. read the letter →

arxiv 2411.09453 v1 pith:N65V7MHR submitted 2024-11-14 cs.CV cs.LG

Chen-Long Duan , Yong Li , Xiu-Shen Wei , Lin Zhao This is my paper · ORCID
classification cs.CVcs.LG
keywords long-tailedobjectdetectionpre-trainingcontrastivelearningdynamicrebalancingdualreconstructionsimplicitybiasLVISrare-classaccuracy
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

Long-tailed object detection fails mostly on rare classes, and the paper argues that the failure starts before fine-tuning: standard pre-training ignores the imbalance and lets models settle for simple patterns. 2DRCL is a pre-training stage that attacks both problems at once, using contrastive learning over whole images and over object proposals, a sampling schedule that progressively favors rare classes at the instance level, and a dual reconstruction task that forces the model to keep fine detail. On LVIS v1.0, the paper reports the highest box average precision among the compared methods for both Faster R-CNN and Mask R-CNN, with the largest gains on rare classes (box $\mathrm{AP}^{bb}_{r}$ rising from 17.5 to 18.6 against the ECM baseline). On balanced COCO the same pre-training also improves box AP to 41.4, so the benefit is not achieved by sacrificing common classes. The claim matters because it makes pre-training, rather than only loss design, a direct and reusable tool for fixing tail-class detection.

What carries the argument

The load-bearing object is the 2DRCL loss, a sum of three terms: the Holistic-Local Contrastive Learning loss (InfoNCE on global views plus InfoNCE on object proposals, with a large queue-based dictionary for negative samples), the Dynamic Rebalancing sampler (a per-class repeat factor built from the harmonic mean of image-level and instance-level occurrence frequencies, with the mixing coefficient $\alpha_d = T/T_{\max}$ growing across epochs), and the Dual Reconstruction loss (pixel-level appearance reconstruction $L_{\mathrm{AR}} = \|x - g(f(x))\|_2^2$ and semantic reconstruction $L_{\mathrm{SR}}$ comparing features of the original image with features of a masked reconstruction). The local contrastive branch is what aligns pre-training with detection; the dynamic rebalancing is what shifts attention to tail classes; the dual reconstruction is what prevents the encoder from discarding the fine details that tail classes need.

What would settle it

A controlled comparison in which the local contrastive positives are formed by matching proposals across the two augmented views using Intersection-over-Union, versus pairing them at random, would settle the mechanism: if random pairing does not lower rare-class box AP on LVIS v1.0 by at least the reported margin over the no-LCL baseline, then the LCL component is not carrying the improvement that the paper attributes to it.

Watch

Extended reading notes

Core claim

The central claim is that a detection-oriented pre-training stage can carry most of the burden in long-tailed object detection, provided it couples three mechanisms. Holistic-Local Contrastive Learning runs an InfoNCE objective both on full augmented views (holistic) and on object proposals produced by a pretrained class-agnostic detector (local), so the detection head, not only the backbone, is pre-trained on object-level features. Dynamic Rebalancing computes, per class, an image-level imbalance score and an instance-level (proposal) imbalance score, folds them into a harmonic mean whose mixing weight shifts from image-level to instance-level as training proceeds, and turns the resulting score into a repeat factor that oversamples rare classes. Dual Reconstruction adds appearance reconstruction (pixel-wise MSE) and semantic reconstruction (consistency between features of the original and of a masked reconstruction), which the authors say counters simplicity bias. The paper's reported evidence is that on LVIS v1.0 this recipe beats both general pre-training baselines and state-of-the-art long-tail losses: with Faster R-CNN R50, box average precision is 27.3 versus 26.7 for the effective class-margins (ECM) loss, and rare-class box AP is 18.6 versus 17.5; with Mask R-CNN R50, mask AP is 27.7 versus 27.4 and rare-class mask AP is 20.4 versus 19.7.

Load-bearing premise

The local contrastive branch assumes that object proposals produced by a pretrained class-agnostic detector on two differently augmented views of the same image can be paired into trustworthy positive pairs, but the paper never specifies the matching rule, so noisy pairing could break the alignment that the method credits for its gains.

Editorial extensions

If this is right

  • If pre-training can shift rare-class accuracy this way, then long-tail methods that redesign the fine-tuning loss are not the only lever; a carefully built pre-training stage can be combined with them for extra gains, as the paper shows with eight existing methods.
  • The dynamic rebalancing schedule implies that the right sampling weight for a class depends on training stage, so fixed resampling ratios that are optimal at the start may be suboptimal later in pre-training.
  • Because the gains appear in Faster R-CNN, Mask R-CNN, and ATSS detectors, the pre-training effect transfers across detector designs rather than fitting one architecture.
  • The equal-epoch comparison (same total training time) indicates that the improvement is a property of the pre-training objective, not merely of longer training, so the method offers a path to better tail performance under a fixed compute budget.
  • The dual reconstruction's two terms each contribute in ablations, suggesting that both pixel fidelity and semantic invariance under masking are needed to defeat simplicity bias for rare classes.

Reading between the lines

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

  • A testable extension would make the local contrastive matching explicit: proposals could be paired across views by IoU, by feature similarity, or by attention, and the choice would likely change rare-class AP; the paper's silence on this point leaves the mechanism underspecified.
  • The dynamic rebalancing idea could transfer to other self-supervised objectives, such as masked image modeling, where sampling of patches or instances is usually static; adapting the repeat factor to training stage might yield similar tail-class gains outside detection.
  • If LCL's effect is to pre-train the detection head, then pushing the same proposals through the regression head (not only the classification head) could extend the gains to localization quality, which the paper reports only indirectly.
  • The simplicity-bias framing suggests an adversarial variant in which the masking ratio for semantic reconstruction grows for rare classes; the paper uses a fixed 25% mask, so adaptivity is an untested natural extension.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes 2DRCL, a pre-training framework for object detection that combines holistic contrastive learning (HCL), local contrastive learning (LCL) on object proposals, a dynamic rebalancing sampler, and a dual reconstruction mechanism (appearance and semantic), together with an unspecified detection loss Ldet. Experiments on COCO and LVIS v1.0 with Faster/Mask R-CNN report gains in overall and rare-class AP over several baselines, including a same-epoch comparison in Table 4b and component ablations in Table 5. The main claim is that 2DRCL provides a transferable pre-training stage that improves long-tailed object detection, especially for tail classes.

Significance. If the protocol issues are resolved, the proposed framework would be a useful pre-training addition to existing long-tailed detectors, with a plausible mechanism (dynamic rebalancing plus reconstruction) and broad empirical coverage: COCO and LVIS, two backbones, two detector heads, plus an ATSS variant and a COCO-LT experiment. The paper also includes component ablations, error analyses, and efficiency measurements. The same-epoch control in Table 4b and the ablation in Table 5 are useful empirical practices. However, the significance is conditional because the central claim depends on the pre-training protocol and the definition of Ldet, which are not specified in the manuscript.

major comments (3)
  1. [Section 4.1/4.2 and Eq. (8)] The pre-training data source for the LVIS experiments is never stated. Section 4.2 says only 'we first use our 2DRCL for pre-training' before fine-tuning, and Appendix A.1 describes the architecture but not the dataset. Moreover, Ldet in Eq. (8) is introduced as 'the loss of object detection' but is never defined anywhere in the paper. If pre-training is performed on LVIS v1.0 train with a supervised detection loss, then the comparisons in Tables 2, 3, A.1, and A.2 pit methods that have not seen target labels during pre-training against a method that has. The same-epoch control in Table 4b equalizes total epochs but not data or label exposure, so it does not remove this confound. Please specify for every experiment: (i) the exact pre-training dataset, (ii) whether ground-truth labels are used, (iii) a precise definition of Ldet, and (iv) the pre-training/fine-tuning epoch split for each table.
  2. [Section 3.1.2, Eq. (2)] The positive proposal zbb+ is said to be 'a corresponding positive object proposal within another augmented input image x+', but the matching rule between the proposals generated independently for x and x+ is never described. Since the class-agnostic detector may output different numbers and locations of boxes per view, the loss is not reproducible without an explicit matching algorithm (e.g., IoU after known geometric transforms, or feature similarity). Please provide the matching procedure and any threshold parameters.
  3. [Table 3(b), R101-FPN row] The claimed rare-class benefit reverses on ResNet-101: APr drops from 21.9 (ECM) to 21.1 (Ours), a 0.8-point decrease, while the overall AP gain is only 0.1. This is inconsistent with the abstract's claim of 'particularly improving the mAP/AP scores for tail classes.' Please discuss this failure case and, more generally, report results over multiple seeds with standard deviations, since all tables report single runs.
minor comments (5)
  1. [Section 3.2, Eq. (4)] The quantity f_c is called the harmonic mean of f_im and f_in, but the denominator is a weighted arithmetic mean, not the standard harmonic mean denominator; please correct the terminology or the formula and clarify the intended weighting.
  2. [Section 3.3.3, after Eq. (8)] The statement 'the weights of all losses in L are set to 1' conflicts with the α_c, β_c, and α_r weights defined in Eqs. (3) and (7); please reconcile these statements.
  3. [Table 1] The 'From scratch' baseline methods use 200-800 pre-training epochs, while Ours uses ImageNet-initialized 12-epoch pre-training; the comparison would be clearer if the total pre-training compute or epochs for each method were listed.
  4. [Table 2] It is not clear which dataset the competing pre-trained methods (MoCo v2, SimCLR, BYOL, SoCo, AlignDet) were pre-trained on before LVIS fine-tuning; please add a column or footnote specifying the pre-training source for each method.
  5. [Appendix A.1] The generator architecture is described, but the encoder architecture used for the reconstruction path is not specified; please clarify how f(x) is obtained from the detection backbone.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity found: the reported AP gains are empirical benchmark results, and the unresolved Ldet and pre-training-dataset questions are reproducibility concerns rather than reductions-by-construction.

full rationale

The derivation chain here is empirical rather than definitional. The proposed pre-training objective in Eq. (8) combines HCL and LCL InfoNCE losses, an RFS-style dynamic repeat factor rc = max(1, sqrt(t/fc)), and AR/SR reconstruction losses; none of the reported COCO/LVIS AP numbers is computed from a fitted parameter or from the method's own definitions. The hyperparameters (alpha_c=0.1, beta_c=0.05, alpha_r=0.1, t=0.001) are tuning choices, not equations that define the reported accuracy. The central evidence is held-out validation performance on COCO and LVIS, so the claims are not equivalent to their inputs by construction. The paper does contain self-citations (e.g., refs. [51,52] for simplicity bias), but that premise is also supported by external references [21,41,44], so no self-citation is load-bearing. Two non-circularity concerns should be flagged explicitly: Eq. (8) introduces Ldet as 'the loss of object detection' without ever defining it, and Section 4.2 states only that 2DRCL is used 'for pre-training, followed by existing long-tailed methods for fine-tuning' without stating the pre-training dataset for the LVIS experiments. If Ldet were a supervised detection loss on the LVIS training split and pre-training used LVIS images, the reported gains could be confounded by extra target-domain supervision; however, the paper does not exhibit the equation or protocol that would make the reported AP reduce to that loss. These are reproducibility and experimental-design risks, not demonstrated circularity.

Assumptions & free parameters 9 free parameters · 4 assumptions · 0 invented entities

The central claim rests on two empirical bets: that proposal-level contrastive pairs are meaningful, and that image-level balancing plus reconstruction changes feature quality for rare classes. The paper provides ablation evidence for the combination but does not independently validate these bets or report sensitivity to the hand-set hyperparameters.

free parameters (9)
  • alpha_c = 0.1
    Weight of the holistic contrastive loss in Eq. 3; chosen by hand, no sensitivity analysis reported.
  • beta_c = 0.05
    Weight of the local contrastive loss in Eq. 3; chosen by hand.
  • alpha_r = 0.1
    Trade-off between appearance and semantic reconstruction in Eq. 7; chosen by hand.
  • t = 0.001
    Threshold in the repeat factor rc = max(1, sqrt(t/fc)); controls how aggressively tail classes are oversampled.
  • temperature tau = 0.2
    Temperature in the InfoNCE losses (Eqs. 1 and 2); standard value, but hand-set.
  • K negatives for HCL = 65536
    Number of negative samples for holistic contrastive learning; chosen by hand.
  • selected proposals per image = 8
    Appendix A.1 states that 8 bounding boxes are randomly selected from the class-agnostic detector for pre-training.
  • SR mask ratio = 25%
    Section 3.3.3 masks a fixed percentage of an object proposal in semantic reconstruction; chosen by hand.
  • alpha_d schedule = T / Tmax
    Linear schedule in Eq. 4 that shifts emphasis from image-level to instance-level balancing; no comparison to other schedules.
assumptions (4)
  • domain assumption The class-agnostic detector from [24] yields high-quality proposals that align pre-training with detection.
    Section 3.1.2 and Appendix A.1; no quantitative validation of proposal quality or matching correctness.
  • ad hoc to paper The harmonic-mean rebalancing score in Eq. 4 with time-varying alpha improves tail representation without overfitting.
    Introduced for this method; the ablation shows DRB helps but does not isolate static versus dynamic versions.
  • domain assumption Pixel reconstruction plus masked feature reconstruction mitigates simplicity bias.
    Section 3.3.1 states this as a hypothesis; evidence is qualitative attention maps and weight-norm trends, not a direct simplicity-bias measurement.
  • standard math MoCo-style InfoNCE and momentum EMA are used as standard, reliable contrastive machinery.
    Eq. 1 and Section 3.1.1; taken from prior work without re-derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Long-Tailed Object Detection Pre-training: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction." pith.science (2026). https://pith.science/paper/N65V7MHR

@misc{pith2026241109453,
  author       = {Pith},
  title        = {Pith review of: Long-Tailed Object Detection Pre-training: Dynamic Rebalancing Contrastive Learning with Dual Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N65V7MHR}},
  note         = {Machine review of arXiv:2411.09453}
}
read the original abstract

Pre-training plays a vital role in various vision tasks, such as object recognition and detection. Commonly used pre-training methods, which typically rely on randomized approaches like uniform or Gaussian distributions to initialize model parameters, often fall short when confronted with long-tailed distributions, especially in detection tasks. This is largely due to extreme data imbalance and the issue of simplicity bias. In this paper, we introduce a novel pre-training framework for object detection, called Dynamic Rebalancing Contrastive Learning with Dual Reconstruction (2DRCL). Our method builds on a Holistic-Local Contrastive Learning mechanism, which aligns pre-training with object detection by capturing both global contextual semantics and detailed local patterns. To tackle the imbalance inherent in long-tailed data, we design a dynamic rebalancing strategy that adjusts the sampling of underrepresented instances throughout the pre-training process, ensuring better representation of tail classes. Moreover, Dual Reconstruction addresses simplicity bias by enforcing a reconstruction task aligned with the self-consistency principle, specifically benefiting underrepresented tail classes. Experiments on COCO and LVIS v1.0 datasets demonstrate the effectiveness of our method, particularly in improving the mAP/AP scores for tail classes.

Figures

Figures reproduced from arXiv: 2411.09453 by the authors.

Figure 1
Figure 1. Illustration of the proposed Dynamic Rebalancing Contrastive Learning with Dual Recon [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Error analyses comparisons. 2DRCL achieves superior performance on tail classes without [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Attention map comparisons w.r.t Baseline [ [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 58 canonical work pages

  1. [1]

    DETReg: Unsupervised pretraining with region priors for object detection

    Amir Bar, Xin Wang, Vadim Kantorov, Colorado J Reed, Roei Herzig, Gal Chechik, Anna Rohrbach, Trevor Darrell, and Amir Globerson. DETReg: Unsupervised pretraining with region priors for object detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 14585–14595, 2022

    work page 2022

  2. [2]

    Unsu- pervised learning of visual features by contrasting cluster assignments

    Mathilde Caron, Ishan Misra, Julien Mairal, Priya Goyal, Piotr Bojanowski, and Armand Joulin. Unsu- pervised learning of visual features by contrasting cluster assignments. In Proc. Adv. Neural Inf. Process. Syst., pages 9912–9924, 2020

    work page 2020

  3. [3]

    MMDetection: Open MMLab detection toolbox and benchmark

    Kai Chen, Jiaqi Wang, Jiangmiao Pang, Yuhang Cao, Yu Xiong, Xiaoxiao Li, Shuyang Sun, Wansen Feng, Ziwei Liu, Jiarui Xu, Zheng Zhang, Dazhi Cheng, Chenchen Zhu, Tianheng Cheng, Qijie Zhao, Buyu Li, Xin Lu, Rui Zhu, Yue Wu, Jifeng Dai, Jingdong Wang, Jianping Shi, Wanli Ouyang, Chen Change Loy, and Dahua Lin. MMDetection: Open MMLab detection toolbox and b...

    arXiv 1906

  4. [4]

    Big self- supervised models are strong semi-supervised learners

    Ting Chen, Simon Kornblith, Kevin Swersky, Mohammad Norouzi, and Geoffrey Hinton. Big self- supervised models are strong semi-supervised learners. arXiv preprint arXiv:2006.10029, 2020

    arXiv 2006

  5. [5]

    Improved baselines with momentum contrastive learning

    Xinlei Chen, Haoqi Fan, Ross Girshick, and Kaiming He. Improved baselines with momentum contrastive learning. arXiv preprint arXiv:2003.04297, 2020

    arXiv 2003

  6. [6]

    Siamese DETR

    Zeren Chen, Gengshi Huang, Wei Li, Jianing Teng, Kun Wang, Jing Shao, Chen Change Loy, and Lu Sheng. Siamese DETR. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 15722–15731, 2023

    work page 2023

  7. [7]

    UP-DETR: Unsupervised pre-training for object detection with transformers

    Zhigang Dai, Bolun Cai, Yugeng Lin, and Junying Chen. UP-DETR: Unsupervised pre-training for object detection with transformers. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 1601–1610, 2021

    work page 2021

  8. [8]

    ImageNet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: A large-scale hierarchical image database. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 248–255, 2009

    work page 2009

Show all 65 references
  1. [9]

    Exploring classification equilibrium in long-tailed object detection

    Chengjian Feng, Yujie Zhong, and Weilin Huang. Exploring classification equilibrium in long-tailed object detection. In Proc. IEEE Int. Conf. Comput. Vis., pages 3417–3426, 2021

    2021

  2. [10]

    Bootstrap your own latent - a new approach to self- supervised learning

    Jean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec, Pierre Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Avila Pires, Zhaohan Guo, Mohammad Gheshlaghi Azar, Bilal Piot, Koray Kavukcuoglu, Remi Munos, and Michal Valko. Bootstrap your own latent - a new ap...

    2020

  3. [11]

    LVIS: A dataset for large vocabulary instance segmentation

    Agrim Gupta, Piotr Dollar, and Ross Girshick. LVIS: A dataset for large vocabulary instance segmentation. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 5356–5364, 2019

    2019

  4. [12]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , pages 15979–15988, 2022

    2022

  5. [13]

    Momentum contrast for unsupervised visual representation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual representation learning. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , pages 9729–9738, 2020

    2020

  6. [14]

    Mask R-CNN

    Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask R-CNN. In Proc. IEEE Int. Conf. Comput. Vis., pages 2961–2969, 2017

    2017

  7. [15]

    Spatial pyramid pooling in deep convolutional networks for visual recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Trans. Pattern Anal. Mach. Intell., 37(9):1904–1916, 2015

    1904

  8. [16]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 770–778, 2016

    2016

  9. [17]

    Relieving long-tailed instance segmentation via pairwise class balance

    Yin-Yin He, Peizhen Zhang, Xiu-Shen Wei, Xiangyu Zhang, and Jian Sun. Relieving long-tailed instance segmentation via pairwise class balance. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , pages 7000–7009, 2022

    2022

  10. [18]

    Efficient visual pretraining with contrastive detection

    Olivier J Hénaff, Skanda Koppula, Jean-Baptiste Alayrac, Aaron van den Oord, Oriol Vinyals, and João Carreira. Efficient visual pretraining with contrastive detection. In Proc. IEEE Int. Conf. Comput. Vis. , pages 10086–10096, 2021

    2021

  11. [19]

    DropLoss for long-tail instance segmentation

    Ting-I Hsieh, Esther Robb, Hwann-Tzong Chen, and Jia-Bin Huang. DropLoss for long-tail instance segmentation. In Proc. AAAI Conf. Artif. Intell., pages 1549–1557, 2021

    2021

  12. [20]

    An asymmetric augmented self-supervised learning method for unsupervised fine-grained image hashing

    Feiran Hu, Chenlin Zhang, Jiangliang Guo, Xiu-Shen Wei, Lin Zhao, Anqi Xu, and Lingyan Gao. An asymmetric augmented self-supervised learning method for unsupervised fine-grained image hashing. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 17648–17657, 2024

    2024

  13. [21]

    The low-rank simplicity bias in deep networks

    Minyoung Huh, Hossein Mobahi, Richard Zhang, Brian Cheung, Pulkit Agrawal, and Phillip Isola. The low-rank simplicity bias in deep networks. arXiv preprint arXiv:2103.10427, 2021

    2021 arXiv

  14. [22]

    Long-tail detection with effective class-margins

    Jang Hyun Cho and Philipp Krähenbühl. Long-tail detection with effective class-margins. In Proc. Eur . Conf. Comput. Vis., pages 698–714, 2022

    2022

  15. [23]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift

    Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In Proc. Int. Conf. Mach. Learn., pages 448–456, 2015

    2015

  16. [24]

    Learning open-world object proposals without learning to classify

    Dahun Kim, Tsung-Yi Lin, Anelia Angelova, In So Kweon, and Weicheng Kuo. Learning open-world object proposals without learning to classify. IEEE Trans. Robot. Autom., 7(2):5453–5460, 2022

    2022

  17. [25]

    AlignDet: Aligning pre-training and fine-tuning in object detection

    Ming Li, Jie Wu, Xionghui Wang, Chen Chen, Jie Qin, Xuefeng Xiao, Rui Wang, Min Zheng, and Xin Pan. AlignDet: Aligning pre-training and fine-tuning in object detection. In Proc. IEEE Int. Conf. Comput. Vis. , pages 6866–6876, 2023

    2023

  18. [26]

    Contrastive learning of person-independent representations for facial action unit detection

    Yong Li and Shiguang Shan. Contrastive learning of person-independent representations for facial action unit detection. IEEE Trans. Image Process., 32:3212–3225, 2023. 11

    2023

  19. [27]

    Overcoming classifier imbalance for long-tail object detection with balanced group softmax

    Yu Li, Tao Wang, Bingyi Kang, Sheng Tang, Chunfeng Wang, Jintao Li, and Jiashi Feng. Overcoming classifier imbalance for long-tail object detection with balanced group softmax. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 10988–10997, 2020

    2020

  20. [28]

    Learning representations for facial actions from unlabeled videos

    Yong Li, Jiabei Zeng, and Shiguang Shan. Learning representations for facial actions from unlabeled videos. IEEE Trans. Pattern Anal. Mach. Intell., 44(1):302–317, 2022

    2022

  21. [29]

    Occlusion aware facial expression recognition using cnn with attention mechanism

    Yong Li, Jiabei Zeng, Shiguang Shan, and Xilin Chen. Occlusion aware facial expression recognition using cnn with attention mechanism. IEEE Trans. Image Process., 28(5):2439–2450, 2019

    2019

  22. [30]

    Self-supervised representation learning from videos for facial action unit detection

    Yong Li, Jiabei Zeng, Shiguang Shan, and Xilin Chen. Self-supervised representation learning from videos for facial action unit detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 10916–10925, 2019

    2019

  23. [31]

    Feature pyramid networks for object detection

    Tsung-Yi Lin, Piotr Dollár, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature pyramid networks for object detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , pages 2117–2125, 2017

    2017

  24. [32]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. Focal loss for dense object detection. In Proc. IEEE Int. Conf. Comput. Vis., pages 2980–2988, 2017

    2017

  25. [33]

    Microsoft COCO: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. Microsoft COCO: Common objects in context. In Proc. Eur . Conf. Comput. Vis., pages 740–755, 2014

    2014

  26. [34]

    Self-EMD: Self-supervised object detection without imagenet

    Songtao Liu, Zeming Li, and Jian Sun. Self-EMD: Self-supervised object detection without imagenet. arXiv preprint arXiv:2011.13677, 2020

    2011 arXiv

  27. [35]

    Rectified linear units improve restricted boltzmann machines

    Vinod Nair and Geoffrey E Hinton. Rectified linear units improve restricted boltzmann machines. In Proc. Int. Conf. Mach. Learn., pages 807–814, 2010

    2010

  28. [36]

    Representation learning with contrastive predictive coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748, 2018

    2018 arXiv

  29. [37]

    Yang, Zach DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Köpf, Edward Z. Yang, Zach DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  30. [38]

    Balanced classification: A unified framework for long-tailed object detection

    Tianhao Qi, Hongtao Xie, Pandeng Li, Jiannan Ge, and Yongdong Zhang. Balanced classification: A unified framework for long-tailed object detection. IEEE Trans. Multimedia, 26:3088–3101, 2024

    2024

  31. [39]

    You only look once: Unified, real-time object detection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 779–788, 2016

    2016

  32. [40]

    Faster R-CNN: Towards real-time object detection with region proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Trans. Pattern Anal. Mach. Intell., 39(6):1137–1149, 2016

    2016

  33. [41]

    The pitfalls of simplicity bias in neural networks

    Harshay Shah, Kaustav Tamuly, Aditi Raghunathan, Prateek Jain, and Praneeth Netrapalli. The pitfalls of simplicity bias in neural networks. In Proc. Adv. Neural Inf. Process. Syst., pages 9573–9585, 2020

    2020

  34. [42]

    Equalization loss v2: A new gradient balance approach for long-tailed object detection

    Jingru Tan, Xin Lu, Gang Zhang, Changqing Yin, and Quanquan Li. Equalization loss v2: A new gradient balance approach for long-tailed object detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , pages 1685–1694, 2021

    2021

  35. [43]

    Equalization loss for long-tailed object recognition

    Jingru Tan, Changbao Wang, Buyu Li, Quanquan Li, Wanli Ouyang, Changqing Yin, and Junjie Yan. Equalization loss for long-tailed object recognition. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 11662–11671, 2020

    2020

  36. [44]

    Evading the simplicity bias: Training a diverse set of models discovers solutions with superior ood generalization

    Damien Teney, Ehsan Abbasnejad, Simon Lucey, and Anton Van den Hengel. Evading the simplicity bias: Training a diverse set of models discovers solutions with superior ood generalization. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 16761–16772, 2022

    2022

  37. [45]

    Seesaw loss for long-tailed instance segmentation

    Jiaqi Wang, Wenwei Zhang, Yuhang Zang, Yuhang Cao, Jiangmiao Pang, Tao Gong, Kai Chen, Ziwei Liu, Chen Change Loy, and Dahua Lin. Seesaw loss for long-tailed instance segmentation. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 9695–9704, 2021

    2021

  38. [46]

    Contrastive learning based hybrid networks for long-tailed image classification

    Peng Wang, Kai Han, Xiu-Shen Wei, Lei Zhang, and Lei Wang. Contrastive learning based hybrid networks for long-tailed image classification. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 943–952, 2021

    2021

  39. [47]

    The devil is in classification: A simple framework for long-tail instance segmentation

    Tao Wang, Yu Li, Bingyi Kang, Junnan Li, Junhao Liew, Sheng Tang, Steven Hoi, and Jiashi Feng. The devil is in classification: A simple framework for long-tail instance segmentation. In Proc. Eur . Conf. Comput. Vis., pages 728–744, 2020

    2020

  40. [48]

    Adaptive class suppression loss for long-tail object detection

    Tong Wang, Yousong Zhu, Chaoyang Zhao, Wei Zeng, Jinqiao Wang, and Ming Tang. Adaptive class suppression loss for long-tail object detection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 3103–3112, 2021

    2021

  41. [49]

    Dense contrastive learning for self-supervised visual pre-training

    Xinlong Wang, Rufeng Zhang, Chunhua Shen, Tao Kong, and Lei Li. Dense contrastive learning for self-supervised visual pre-training. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 3024–3033, 2021

    2021

  42. [50]

    Aligning pretraining for detection via object-level contrastive learning

    Fangyun Wei, Yue Gao, Zhirong Wu, Han Hu, and Stephen Lin. Aligning pretraining for detection via object-level contrastive learning. In Proc. Adv. Neural Inf. Process. Syst., pages 22682–22694, 2021

    2021

  43. [51]

    Attribute-aware deep hashing with self-consistency for large-scale fine-grained image retrieval

    Xiu-Shen Wei, Yang Shen, Xuhao Sun, Peng Wang, and Yuxin Peng. Attribute-aware deep hashing with self-consistency for large-scale fine-grained image retrieval. IEEE Trans. Pattern Anal. Mach. Intell. , 45(11):13904–13920, 2023. 12

    2023

  44. [52]

    Delving deep into simplicity bias for long-tailed image recognition

    Xiu-Shen Wei, Xuhao Sun, Yang Shen, Anqi Xu, Peng Wang, and Faen Zhang. Delving deep into simplicity bias for long-tailed image recognition. arXiv preprint arXiv:2302.03264, 2023

    2023 arXiv

  45. [53]

    Negatives make a positive: An embarrassingly simple approach to semi-supervised few-shot learning

    Xiu-Shen Wei, He-Yang Xu, Zhiwen Yang, Chen-Long Duan, and Yuxin Peng. Negatives make a positive: An embarrassingly simple approach to semi-supervised few-shot learning. IEEE Trans. Pattern Anal. Mach. Intell., 46(4):2091–2103, 2024

    2024

  46. [54]

    Prototype-based classifier learning for long-tailed visual recognition

    Xiu-Shen Wei, Shu-Lin Xu, Hao Chen, Liang Xiao, and Yuxin Peng. Prototype-based classifier learning for long-tailed visual recognition. Sci. China Inf. Sci. , 65(6):160105, 2022

    2022

  47. [55]

    Self-supervised visual representation learning with semantic grouping

    Xin Wen, Bingchen Zhao, Anlin Zheng, Xiangyu Zhang, and Xiaojuan Qi. Self-supervised visual representation learning with semantic grouping. In Proc. Adv. Neural Inf. Process. Syst., pages 16423– 16438, 2022

    2022

  48. [56]

    Instance-aware repeat factor sampling for long-tailed object detection

    Burhaneddin Yaman, Tanvir Mahmud, and Chun-Hao Liu. Instance-aware repeat factor sampling for long-tailed object detection. arXiv preprint arXiv:2305.08069, 2023

    2023 arXiv

  49. [57]

    A survey on long-tailed visual recognition

    Lu Yang, He Jiang, Qing Song, and Jun Guo. A survey on long-tailed visual recognition. Int. J. Comput. Vis., 130(7):1837–1872, 2022

    2022

  50. [58]

    Reconciling object-level and global-level objectives for long-tail detection

    Shaoyu Zhang, Chen Chen, and Silong Peng. Reconciling object-level and global-level objectives for long-tail detection. In Proc. IEEE Int. Conf. Comput. Vis., pages 18982–18992, 2023

    2023

  51. [59]

    Shifeng Zhang, Cheng Chi, Yongqiang Yao, Zhen Lei, and Stan Z. Li. Bridging the gap between anchor- based and anchor-free detection via adaptive training sample selection. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 9759–9768, 2020

    2020

  52. [60]

    Self-supervised aggregation of diverse experts for test-agnostic long-tailed recognition

    Yifan Zhang, Bryan Hooi, Lanqing Hong, and Jiashi Feng. Self-supervised aggregation of diverse experts for test-agnostic long-tailed recognition. In Proc. Adv. Neural Inf. Process. Syst. , pages 34077–34090, 2022

    2022

  53. [61]

    Deep long-tailed learning: A survey

    Yifan Zhang, Bingyi Kang, Bryan Hooi, Shuicheng Yan, and Jiashi Feng. Deep long-tailed learning: A survey. IEEE Trans. Pattern Anal. Mach. Intell., 45(9):10795–10816, 2023

    2023

  54. [62]

    BBN: Bilateral-branch network with cumulative learning for long-tailed visual recognition

    Boyan Zhou, Quan Cui, Xiu-Shen Wei, and Zhao-Min Chen. BBN: Bilateral-branch network with cumulative learning for long-tailed visual recognition. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 9719–9728, 2020

    2020

  55. [63]

    Generalized logit adjustment: Calibrating fine-tuned models by removing label bias in foundation models

    Beier Zhu, Kaihua Tang, Qianru Sun, and Hanwang Zhang. Generalized logit adjustment: Calibrating fine-tuned models by removing label bias in foundation models. In Proc. Adv. Neural Inf. Process. Syst., pages 64663–64680, 2023

    2023

  56. [64]

    Balanced contrastive learning for long-tailed visual recognition

    Jianggang Zhu, Zheng Wang, Jingjing Chen, Yi-Ping Phoebe Chen, and Yu-Gang Jiang. Balanced contrastive learning for long-tailed visual recognition. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 6898–6907, 2022. 13 A Appendix / supplemental material In the supplemen...

    2022

  57. [256]

    Once all views are constructed, we employ the data augmentation pipeline of MoCo [ 5, 13]. Our generator architecture consists of four deconvolutional layers with dimensions (2048, 512), (512, 256), (256, 64), and (64, 3), respectively, and each layer uses a kernel size of 4. ...

    2017

Pith tools

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