Long-tailed Recognition by Routing Diverse Distribution-Aware Experts

Wang, Xudong; Lian, Long; Miao, Zhongqi; Liu, Ziwei; Yu, Stella X.

doi:10.48550/arxiv.2010.01809

Cited by 51 publications

(89 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent works proposed decoupling training [18,48] which first obtains a good representation through conventional training and then calibrates the classifier. Other techniques like ensemble [40], self-supervised learning [20,44] and knowledge distillation [15,20] are verified to be useful in long-tailed learning.…”

Section: Related Workmentioning

confidence: 99%

Relieving Long-tailed Instance Segmentation via Pairwise Class Balance

He¹,

Zhang²,

Wei³

et al. 2022

Preprint

View full text Add to dashboard Cite

Long-tailed instance segmentation is a challenging task due to the extreme imbalance of training samples among classes. It causes severe biases of the head classes (with majority samples) against the tailed ones. This renders "how to appropriately define and alleviate the bias" one of the most important issues. Prior works mainly use label distribution or mean score information to indicate a coarse-grained bias. In this paper, we explore to excavate the confusion matrix, which carries the fine-grained misclassification details, to relieve the pairwise biases, generalizing the coarse one. To this end, we propose a novel Pairwise Class Balance (PCB) method, built upon a confusion matrix which is updated during training to accumulate the ongoing prediction preferences. PCB generates fightback soft labels for regularization during training. Besides, an iterative learning paradigm is developed to support a progressive and smooth regularization in such debiasing. PCB can be plugged and played to any existing method as a complement. Experimental results on LVIS demonstrate that our method achieves state-of-theart performance without bells and whistles. Superior results across various architectures show the generalization ability.

show abstract

Section: Related Workmentioning

confidence: 99%

Relieving Long-tailed Instance Segmentation via Pairwise Class Balance

He¹,

Zhang²,

Wei³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Each classifier focuses on the classification of a small and relatively balanced group of classes from the data. [40] described a shared architecture for multiple classifiers, a distribution-aware loss, and an expert routing module. The current paper proposes a two-stage approach for training one model with a single classifier.…”

Section: Long-tail Recognitionmentioning

confidence: 99%

Distributional Robustness Loss for Long-tail Learning

Samuel

Chechik

2021

Preprint

View full text Add to dashboard Cite

Real-world data is often unbalanced and long-tailed, but deep models struggle to recognize rare classes in the presence of frequent classes. To address unbalanced data, most studies try balancing the data, the loss, or the classifier to reduce classification bias towards head classes. Far less attention has been given to the latent representations learned with unbalanced data. We show that the feature extractor part of deep networks suffers greatly from this bias. We propose a new loss based on robustness theory, which encourages the model to learn high-quality representations for both head and tail classes. While the general form of the robustness loss may be hard to compute, we further derive an easy-to-compute upper bound that can be minimized efficiently. This procedure reduces representation bias towards head classes in the feature space and achieves new SOTA results on CIFAR100-LT, ImageNet-LT, and iNaturalist long-tail benchmarks. We find that training with robustness increases recognition accuracy of tail classes while largely maintaining the accuracy of head classes. The new robustness loss can be combined with various classifier balancing techniques and can be applied to representations at several layers of the deep model.

show abstract

“…𝜏-norm [10] OLTR [15] LDAM+DRW [1] cRT [10] Progressive resampling + IM CB Resampling +IM CB Focal Loss + IM Input Mix-up [30] Class-balanced Cross-entropy Loss Focal Loss [14] Progressive Resampling [10] Class-balanced Resampling [9] ReMix [2] Manifold Mix-up [30] BBN [34] One-stage: re-balancing and augmentations Multi-stage: multi-stage training and transfer learning RIDE [23] LFME [26] Figure 1. Performance of representative long-tailed recognition methods in terms of majority and minority classes compared to the baseline model (a ResNet).…”

Section: Baseline Ace (Ours)mentioning

confidence: 99%

ACE: Ally Complementary Experts for Solving Long-Tailed Recognition in One-Shot

Cai

Wang

Hwang

2021

Preprint

View full text Add to dashboard Cite

One-stage long-tailed recognition methods improve the overall performance in a "seesaw" manner, i.e., either sacrifice the head's accuracy for better tail classification or elevate the head's accuracy even higher but ignore the tail. Existing algorithms bypass such trade-off by a multi-stage training process: pre-training on imbalanced set and finetuning on balanced set. Though achieving promising performance, not only are they sensitive to the generalizability of the pre-trained model, but also not easily integrated into other computer vision tasks like detection and segmentation, where pre-training of classifiers solely is not applicable. In this paper, we propose a one-stage long-tailed recognition scheme, ally complementary experts (ACE), where the expert is the most knowledgeable specialist in a subset that dominates its training, and is complementary to other experts in the less-seen categories without being disturbed by what it has never seen. We design a distributionadaptive optimizer to adjust the learning pace of each expert to avoid over-fitting. Without special bells and whistles, the vanilla ACE outperforms the current one-stage SOTA method by 3∼ 10% on CIFAR10-LT, CIFAR100-LT, ImageNet-LT and iNaturalist datasets. It is also shown to be the first one to break the "seesaw" trade-off by improving the accuracy of the majority and minority categories simultaneously in only one stage. Code and trained models are at https://github.com/jrcai/ACE.

show abstract

Long-tailed Recognition by Routing Diverse Distribution-Aware Experts

Cited by 51 publications

References 28 publications

Relieving Long-tailed Instance Segmentation via Pairwise Class Balance

Relieving Long-tailed Instance Segmentation via Pairwise Class Balance

Distributional Robustness Loss for Long-tail Learning

ACE: Ally Complementary Experts for Solving Long-Tailed Recognition in One-Shot

Contact Info

Product

Resources

About