Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation

By Hung-Yu Tseng, Hsin-Ying Lee, Jia-Bin Huang, Ming-Hsuan Yang. 2020

479 citations recorded by the source corpus.

Few-shot classification aims to recognize novel categories with only few labeled images in each class. Existing metric-based few-shot classification algorithms predict categories by comparing the feature embeddings of query images with those from a few labeled images (support examples) using a learned metric function. While promising performance has been demonstrated, these methods often fail to generalize to unseen domains due to large discrepancy of the feature distribution across domains. In this work, we address the problem of few-shot classification under domain shifts for metric-based methods. Our core idea is to use feature-wise transformation layers for augmenting the image features using affine transforms to simulate various feature distributions under different domains in the training stage. To capture variations of the feature distributions under different domains, we further apply a learning-to-learn approach to search for the hyper-parameters of the feature-wise transformation layers. We conduct extensive experiments and ablation studies under the domain generalization setting using five few-shot classification datasets: mini-ImageNet, CUB, Cars, Places, and Plantae. Experimental results demonstrate that the proposed feature-wise transformation layer is applicable to various metric-based models, and provides consistent improvements on the few-shot classification performance under domain shift.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

In Representation & Self-Supervised Learning

Citations in the corpus

Repurposing Pretrained Models for Robust Out-of-domain Few-Shot Learning (2021, methodology): “Tseng et al. (2020) used a randomly split manner for the cross-domain test and Triantafillou et al. (2020) used all traffic signs dataset for the test.”
Meta-Learning with Less Forgetting on Large-Scale Non-Stationary Task Distributions (2022, neutral): “We follow the split of [55] and split the dataset into 100/50/50 classes for meta-training/validation/test.”
Meta-Learning in Neural Networks: A Survey (2020, background): “…[168]–[172] Dataset/Environment [144] [146] [149] [150] Instance Weights [142] Feature/Metric [20], [86]–[88] Data Augmentation/Noise Generator [134] [173] [174] [133] [135] Modules [130], [131] Annotation Policy [175], [176] [177] TABLE 1 Classiﬁcation of research papers according to our taxonomy.”
Regularizing Generative Adversarial Networks under Limited Data (2021, methodology): “Recent methods have shown successes in overcoming training data bias such as label noise [3, 17, 25, 26, 43, 54, 57, 77] and biased data distributions [9, 11, 38, 41, 61, 69, 71].”
A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities (2022, methodology): “FWT [85] utilizes only the source data for the affine transformation of features, as do LRP-GNN [86] and SBMTL [87].”
Few-Shot Learning With a Strong Teacher (2021, methodology): “We use [111] as the basic GCN-based algorithm (denoted as “GCN”), following [112].”
Learning a Universal Template for Few-shot Dataset Generalization (2021, background): “…< eleni@cs.toronto.edu > . work is shifting towards a more challenging instance of few-shot classiﬁcation across datasets (Chen et al., 2019; Tseng et al., 2020; Triantaﬁllou et al., 2020; Requeima et al., 2019; Bateni et al., 2020; Bronskill et al., 2020; Saikia et al., 2020; Dvornik et…”
StyleMeUp: Towards Style-Agnostic Sketch-Based Image Retrieval (2021, background): “Among them, those designed for domain adaptation [53] or domain generalization [26, 2] are the most relevant to our work.”
Meta Learning on a Sequence of Imbalanced Domains with Difficulty Awareness (2021, background): “Multi-domain meta learning [53, 54, 58] assume tasks from all domains are available during meta training.”
Revisiting Mid-Level Patterns for Cross-Domain Few-Shot Recognition (2020, unknown): “To solve this problem, very recently, cross-domain FSL (CDFSL) [5, 40] has been proposed to handle a more realistic setting where data from the general domain (which is easier to collect [5], e.”
A Broader Study of Cross-Domain Few-Shot Learning (2019, methodology): “All established works in this space have built crossdomain evaluation benchmarks that are limited to natural images [58,5,56].”