Improving Long-Tailed Classification from Instance Level

• URL: http://arxiv.org/abs/2104.06094v1
• Date: Tue, 13 Apr 2021 11:00:19 GMT
• Title: Improving Long-Tailed Classification from Instance Level
• Authors: Yan Zhao, Weicong Chen, Xu Tan, Kai Huang, Jin Xu, Changhu Wang, and Jihong Zhu
• Abstract summary: We propose two instance-level components to improve long-tailed classification. The first is an Adaptive Logit Adjustment (ALA) loss, which applies an adaptive adjusting term to the logit. The second is a Mixture-of-Experts (MoE) network, which contains a multi-expert module and an instance-aware routing module.
• Score: 34.10943893320389
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Data in the real world tends to exhibit a long-tailed label distribution, which poses great challenges for neural networks in classification. Existing methods tackle this problem mainly from the coarse-grained class level, ignoring the difference among instances, e.g., hard samples vs. easy samples. In this paper, we revisit the long-tailed problem from the instance level and propose two instance-level components to improve long-tailed classification. The first one is an Adaptive Logit Adjustment (ALA) loss, which applies an adaptive adjusting term to the logit. Different from the adjusting terms in existing methods that are class-dependent and only focus on tail classes, we carefully design an instance-specific term and add it on the class-dependent term to make the network pay more attention to not only tailed class, but more importantly hard samples. The second one is a Mixture-of-Experts (MoE) network, which contains a multi-expert module and an instance-aware routing module. The routing module is designed to dynamically integrate the results of multiple experts according to each input instance, and is trained jointly with the experts network in an end-to-end manner.Extensive experiment results show that our method outperforms the state-of-the-art methods by 1% to 5% on common long-tailed benchmarks including ImageNet-LT and iNaturalist.

Related papers

• Adapted-MoE: Mixture of Experts with Test-Time Adaption for Anomaly Detection [10.12283550685127]
  We propose an Adapted-MoE to handle multiple distributions of same-category samples by divide and conquer. Specifically, we propose a routing network based on representation learning to route same-category samples into the subclasses feature space. We propose the test-time adaption to eliminate the bias between the unseen test sample representation and the feature distribution learned by the expert model.
  arXiv  Detail & Related papers  (2024-09-09T13:49:09Z)
• Balancing Logit Variation for Long-tailed Semantic Segmentation [28.92929059563813]
  We introduce category-wise variation into the network predictions in the training phase. We close the gap between the feature areas of different categories, resulting in a more balanced representation. Our method manifests itself in strong generalizability to various datasets and task settings.
  arXiv  Detail & Related papers  (2023-06-03T09:19:24Z)
• Relieving Long-tailed Instance Segmentation via Pairwise Class Balance [85.53585498649252]
  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. 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.
  arXiv  Detail & Related papers  (2022-01-08T07:48:36Z)
• Adaptive Class Suppression Loss for Long-Tail Object Detection [49.7273558444966]
  We devise a novel Adaptive Class Suppression Loss (ACSL) to improve the detection performance of tail categories. Our ACSL achieves 5.18% and 5.2% improvements with ResNet50-FPN, and sets a new state of the art.
  arXiv  Detail & Related papers  (2021-04-02T05:12:31Z)
• Train a One-Million-Way Instance Classifier for Unsupervised Visual Representation Learning [45.510042484456854]
  This paper presents a simple unsupervised visual representation learning method with a pretext task of discriminating all images in a dataset using a parametric, instance-level computation. The overall framework is a replica of a supervised classification model, where semantic classes (e.g., dog, bird, and ship) are replaced by instance IDs. scaling up the classification task from thousands of semantic labels to millions of instance labels brings specific challenges including 1) the large-scale softmax classifier; 2) the slow convergence due to the infrequent visiting of instance samples; and 3) the massive number of negative classes that can be noisy.
  arXiv  Detail & Related papers  (2021-02-09T14:44:18Z)
• The Devil is in Classification: A Simple Framework for Long-tail Object Detection and Instance Segmentation [93.17367076148348]
  We investigate performance drop of the state-of-the-art two-stage instance segmentation model Mask R-CNN on the recent long-tail LVIS dataset. We unveil that a major cause is the inaccurate classification of object proposals. We propose a simple calibration framework to more effectively alleviate classification head bias with a bi-level class balanced sampling approach.
  arXiv  Detail & Related papers  (2020-07-23T12:49:07Z)
• Overcoming Classifier Imbalance for Long-tail Object Detection with Balanced Group Softmax [88.11979569564427]
  We provide the first systematic analysis on the underperformance of state-of-the-art models in front of long-tail distribution. We propose a novel balanced group softmax (BAGS) module for balancing the classifiers within the detection frameworks through group-wise training. Extensive experiments on the very recent long-tail large vocabulary object recognition benchmark LVIS show that our proposed BAGS significantly improves the performance of detectors.
  arXiv  Detail & Related papers  (2020-06-18T10:24:26Z)
• Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition from a Domain Adaptation Perspective [98.70226503904402]
  Object frequency in the real world often follows a power law, leading to a mismatch between datasets with long-tailed class distributions. We propose to augment the classic class-balanced learning by explicitly estimating the differences between the class-conditioned distributions with a meta-learning approach.
  arXiv  Detail & Related papers  (2020-03-24T11:28:42Z)
• Few-Shot Learning as Domain Adaptation: Algorithm and Analysis [120.75020271706978]
  Few-shot learning uses prior knowledge learned from the seen classes to recognize the unseen classes. This class-difference-caused distribution shift can be considered as a special case of domain shift. We propose a prototypical domain adaptation network with attention (DAPNA) to explicitly tackle such a domain shift problem in a meta-learning framework.
  arXiv  Detail & Related papers  (2020-02-06T01:04:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.