Harnessing the Power of Local Representations for Few-Shot Classification
Not on the map. Why?
Abstract (arXiv)
Generalizing to novel classes unseen during training is a key challenge of few-shot classification. Recent metric-based methods try to address this by local representations. However, they are unable to take full advantage of them due to (i) improper supervision for pretraining the feature extractor, and (ii) lack of adaptability in the metric for handling various possible compositions of local feature sets. In this work, we harness the power of local representations in improving novel-class generalization. For the feature extractor, we design a novel pretraining paradigm that learns randomly cropped patches by soft labels. It utilizes the class-level diversity of patches while diminishing the impact of their semantic misalignments to hard labels. To align network output with soft labels, we also propose a UniCon KL-Divergence that emphasizes the equal contribution of each base class in describing "non-base" patches. For the metric, we formulate measuring local feature sets as an entropy-regularized optimal transport problem to introduce the ability to handle sets consisting of homogeneous elements. Furthermore, we design a Modulate Module to endow the metric with the necessary adaptability. Our method achieves new state-of-the-art performance on three popular benchmarks. Moreover, it exceeds state-of-the-art transductive and cross-modal methods in the fine-grained scenario.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 38 of its 58 references that are in the map. First posted on arXiv 2 Jul 2024.
Builds on
38 of 58 references.
References · 38
- 2017Attention Is All You Needcited by 59,932
- 2015Distilling the Knowledge in a Neural Networkcited by 10,076
- 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
- 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,482
- 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
- 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,712
- 2017Prototypical Networks for Few-shot Learningcited by 3,001
- 2013Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distancescited by 2,277
- 2016Matching Networks for One Shot Learningcited by 2,246
- 2017Learning to Compare: Relation Network for Few-Shot Learningcited by 1,139
- 2019A Closer Look at Few-shot Classificationcited by 679
- 2018Born Again Neural Networkscited by 539
26 more
- 2013Do Deep Nets Really Need to be Deep?cited by 466
- 2018Meta-Learning for Semi-Supervised Few-Shot Classificationcited by 456
- 2018TADAM: Task dependent adaptive metric for improved few-shot learningcited by 408
- 2020Noise or Signal: The Role of Image Backgrounds in Object Recognitioncited by 342
- 2019A Comprehensive Overhaul of Feature Distillationcited by 341
- 2018Few-Shot Learning via Embedding Adaptation with Set-to-Set Functionscited by 293
- 2018Paraphrasing Complex Network: Network Compression via Factor Transfercited by 255
- 2020Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learningcited by 179
- 2022Pushing the Limits of Simple Pipelines for Few-Shot Learning: External Data and Fine-Tuning Make a Differencecited by 123
- 2019Revisiting Local Descriptor based Image-to-Class Measure for Few-shot Learningcited by 121
- 2020Few-Shot Classification with Feature Map Reconstruction Networkscited by 116
- 2020Interventional Few-Shot Learningcited by 112
- 2018Delta-encoder: an effective sample synthesis method for few-shot object recognitioncited by 89
- 2020Self-Distillation as Instance-Specific Label Smoothingcited by 71
- 2019Associative Alignment for Few-shot Image Classificationcited by 58
- 2017Few-Shot Learning Through an Information Retrieval Lenscited by 51
- 2020DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learningcited by 49
- 2021Partner-Assisted Learning for Few-Shot Image Classificationcited by 36
- 2022Rethinking Generalization in Few-Shot Classificationcited by 36
- 2023M3Net: Multi-view Encoding, Matching, and Fusion for Few-shot Fine-grained Action Recognitioncited by 34
- 2019LGM-Net: Learning to Generate Matching Networks for Few-Shot Learningcited by 27
- 2021Transductive Few-Shot Classification on the Oblique Manifoldcited by 23
- 2021Learning to Affiliate: Mutual Centralized Learning for Few-shot Classificationcited by 18
- 2020Attributes-Guided and Pure-Visual Attention Alignment for Few-Shot Recognitioncited by 15
- 2021ECKPN: Explicit Class Knowledge Propagation Network for Transductive Few-shot Learningcited by 11
- 2022Few-shot Fine-grained Image Classification via Multi-Frequency Neighborhood and Double-cross Modulationcited by 4
Built on by
Citing papers per year: 2026: 1.
1 paper.