mlsanityData to 14 Oct 2026

Channel Importance Matters in Few-Shot Image Classification

arXiv 2206.08126 ↗First posted 16 Jun 2022cs.CVcited by 20

Not on the map. Why?

Abstract (arXiv)

Few-Shot Learning (FSL) requires vision models to quickly adapt to brand-new classification tasks with a shift in task distribution. Understanding the difficulties posed by this task distribution shift is central to FSL. In this paper, we show that a simple channel-wise feature transformation may be the key to unraveling this secret from a channel perspective. When facing novel few-shot tasks in the test-time datasets, this transformation can greatly improve the generalization ability of learned image representations, while being agnostic to the choice of training algorithms and datasets. Through an in-depth analysis of this transformation, we find that the difficulty of representation transfer in FSL stems from the severe channel bias problem of image representations: channels may have different importance in different tasks, while convolutional neural networks are likely to be insensitive, or respond incorrectly to such a shift. This points out a core problem of the generalization ability of modern vision systems and needs further attention in the future. Our code is available at https://github.com/Frankluox/Channel_Importance_FSL.

Key facts. Cited by 20 papers in the map (data through 14 Oct 2026). Builds on 43 of its 53 references that are in the map. First posted on arXiv 16 Jun 2022.

Builds on

43 of 53 references.

References · 43

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  3. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  4. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  5. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,897
  6. 2017Squeeze-and-Excitation Networkscited by 4,789
  7. 2016Wide Residual Networkscited by 3,070
  8. 2017Prototypical Networks for Few-shot Learningcited by 3,015
  9. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  10. 2016Matching Networks for One Shot Learningcited by 2,255
  11. 2018Moment Matching for Multi-Source Domain Adaptationcited by 1,331
  12. 2019Big Transfer (BiT): General Visual Representation Learningcited by 769
31 more
  1. 2017Network Dissection: Quantifying Interpretability of Deep Visual Representationscited by 708
  2. 2019A Closer Look at Few-shot Classificationcited by 684
  3. 2019Meta-Learning with Implicit Gradientscited by 525
  4. 2018Meta-Learning with Latent Embedding Optimizationcited by 476
  5. 2019Meta-Learning with Differentiable Convex Optimizationcited by 461
  6. 2020Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need?cited by 437
  7. 2018TADAM: Task dependent adaptive metric for improved few-shot learningcited by 408
  8. 2018Meta-learning with differentiable closed-form solverscited by 397
  9. 2019Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAMLcited by 359
  10. 2019Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examplescited by 315
  11. 2019A Baseline for Few-Shot Image Classificationcited by 298
  12. 2019Cross Attention Network for Few-shot Classificationcited by 219
  13. 2020Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformationcited by 185
  14. 2014Object Detectors Emerge in Deep Scene CNNscited by 179
  15. 2020Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learningcited by 179
  16. 2020CrossTransformers: spatially-aware few-shot transfercited by 174
  17. 2021Free Lunch for Few-shot Learning: Distribution Calibrationcited by 165
  18. 2019Charting the Right Manifold: Manifold Mixup for Few-shot Learningcited by 136
  19. 2019Fast and Flexible Multi-Task Classification Using Conditional Neural Adaptive Processescited by 131
  20. 2020Negative Margin Matters: Understanding Margin in Few-shot Classificationcited by 129
  21. 2021Cross-domain Few-shot Learning with Task-specific Adapterscited by 80
  22. 2019Meta-Curvaturecited by 70
  23. 2021Learning a Universal Template for Few-shot Dataset Generalizationcited by 61
  24. 2021Exploring Complementary Strengths of Invariant and Equivariant Representations for Few-Shot Learningcited by 56
  25. 2021Rectifying the Shortcut Learning of Background for Few-Shot Learningcited by 51
  26. 2020Fine-grained Angular Contrastive Learning with Coarse Labelscited by 30
  27. 2021How to Train Your MAML to Excel in Few-Shot Classificationcited by 22
  28. 2021On sensitivity of meta-learning to support datacited by 15
  29. 2021Boosting Few-Shot Classification with View-Learnable Contrastive Learningcited by 11
  30. 2020Towards Cross-Granularity Few-Shot Learning: Coarse-to-Fine Pseudo-Labeling with Visual-Semantic Meta-Embeddingcited by 7
  31. 2021Exploring Category-correlated Feature for Few-shot Image Classificationcited by 4

Built on by

Cited by year

Citing papers per year: 2022: 2; 2023: 11; 2024: 5; 2025: 2.

20 papers.

Sort:

  1. 2023A Closer Look at Few-shot Classification Againcited by 46
  2. 2022Alleviating the Sample Selection Bias in Few-shot Learning by Removing Projection to the Centroidcited by 20
  3. 2023SCT: A Simple Baseline for Parameter-Efficient Fine-Tuning via Salient Channelscited by 20
  4. 2023Boosting Object Detection with Zero-Shot Day-Night Domain Adaptationcited by 17
  5. 2024Lightweight Frequency Masker for Cross-Domain Few-Shot Semantic Segmentationcited by 17
  6. 2023A Closer Look at Few-Shot 3D Point Cloud Classificationcited by 10
  7. 2023From Channel Bias to Feature Redundancy: Uncovering the "Less is More" Principle in Few-Shot Learningcited by 10
  8. 2023Taxonomy Adaptive Cross-Domain Adaptation in Medical Imaging via Optimization Trajectory Distillationcited by 7
  9. 2023Mitigating the Effect of Incidental Correlations on Part-based Learningcited by 6
  10. 2023Revisit Parameter-Efficient Transfer Learning: A Two-Stage Paradigmcited by 5
  11. 2023DiffKendall: A Novel Approach for Few-Shot Learning with Differentiable Kendall's Rank Correlationcited by 4
  12. 2023Cross-Level Distillation and Feature Denoising for Cross-Domain Few-Shot Classificationcited by 3
8 more
  1. 2023Towards Redundancy-Free Sub-networks in Continual Learningcited by 3
  2. 2024Channel-Selective Normalization for Label-Shift Robust Test-Time Adaptationcited by 3
  3. 2024A Benchmark Grocery Dataset of Realworld Point Clouds From Single Viewcited by 2
  4. 2024FeTT: Continual Class Incremental Learning via Feature Transformation Tuningcited by 1
  5. 2025Contrastive Graph Modeling for Cross-Domain Few-Shot Medical Image Segmentationcited by 1
  6. 2022Class-Specific Channel Attention for Few-Shot Learningcited by 0
  7. 2024Like Humans to Few-Shot Learning through Knowledge Permeation of Vision and Textcited by 0
  8. 2025XTransfer: Modality-Agnostic Few-Shot Model Transfer for Human Sensing at the Edgecited by 0