不确定性域感知网络在少样本跨域图像分类中的研究  

Uncertainty domain awareness network for cross-domain few-shot image classification

作　　者：余悦 陈楠 成科扬[1] Yu Yue;Chen Nan;Cheng Keyang(School of Computer Science and Communication Engineering,Jiangsu University,Zhenjiang 212013,China)

机构地区：[1]江苏大学计算机科学与通信工程学院,镇江212013

出　　处：《中国图象图形学报》2025年第2期518-532,共15页Journal of Image and Graphics

基　　金：国家自然科学基金项目(62372215);江苏省重点研发计划资助(BE2022781)。

摘　　要：目的跨域少样本学习的主要挑战在于,很难将源域的知识推广到未知的目标域中。最近的一些少样本学习模型试图通过在元训练过程中诱导图像多样化来解决这一问题。然而,其中一些模拟未知领域任务的方法效果有限,因为它们不能有效地模拟领域偏移,其生成的内容变化范围狭窄,难以从域偏移中学习到有效的域不变知识。为了提升少样本模型的跨域泛化能力,提出了一个基于不确定性增强的域感知网络(uncertainty enhancement based domain-aware network,UEDA)。方法基于不确定性增强的域感知框架从特征通道视角探索和提取其中可用于缓解领域偏移的关键知识。首先提出一个不确定性特征增强方法,将特征的充分统计定值定义为服从高斯分布的概率表示,以源域充分统计量为分布中心建模不确定性分布。随后,从不确定性分布中生成有别于加性扰动的挑战性特征,从而挖掘不同域之间的共性知识;其次,提出了基于不确定性增强的域感知方法,将源特征和生成特征视为来自不同领域的信息,利用域鉴别器计算特征通道与领域信息的相关性,从而帮助模型挖掘领域之间的潜在关联并鉴别出其中的域因果信息用于学习。结果实验使用Mini-ImageNet、CUB(caltech-ucsd birds)、Plantae、EuroSAT(land use and land cover classification with sentinel-2)和Cropdiseases共5个数据集评估所提出方法的跨域泛化表现。实验遵从纯源域泛化,其中在图神经网络(graph neural network,GNN)分类框架下,以Mini-ImageNet数据集作为源域,模型在后4个目标域的1-shot和5-shot设置下其平均精度分别为59.50%、47.48%、79.04%和75.08%,表明了所提出方法能有效提高基于源域的跨域图像分类能力。结论本文所提出的基于不确定性增强的域感知网络框架使得模型在训练阶段适应各种域偏移,并从中学习到有效的可泛化知识,从而提高在少样本条件�Objective Inspired by the fast learning properties of humans and transfer learning,researchers have proposed a new machine learning paradigm——few-shot learning.The purpose of few-shot learning is to enable models to quickly learn new knowledge or concepts from a limited amount of labeled data that can be used in unknown categories.Currently,few-shot image classification is based on the framework of meta-learning,which divides the model learning process into a meta-training phase and a meta-testing phase.Existing solutions can be broadly classified according to the following differ⁃ences in concepts:1)optimization-based methods,the basic idea of which is to allow the model to find the parameters that can optimize performance under the training of multiple sets of task data;2)metric learning-based methods,whose core idea is to construct an optimal embedding space for measuring distances so that the distance between similar samples is as small as possible;and 3)data manipulation-based methods,which use some basic data augmentation(e.g.,rotating,clipping,and adding noise)to increase the diversity of the training samples and the amount of data in these three main cat⁃egories.However,these works tend to follow strict assumptions,such as the smoothness assumption,clustering assump⁃tion,and prevalence assumption,and require that the training data and test data come from the same distribution.This situation makes it difficult to ensure data from the same distribution setting during the training process of the model in cer⁃tain real-world scenarios,such as medical imaging,military applications,and finance,where issues such as difficulty in data access and data privacy make it challenging to use labeled data from other domains to provide a priori knowledge.Here,an uncertainty enhancement-based domain-awareness network(UEDA)is proposed to alleviate the problem of domain distribution bias encountered in the learning process of the few-shot model.Method The uncertainty enhancementbased domain-awareness approach exp

关 键 词：图像分类 跨域少样本图像分类 少样本学习(FSL) 域泛化 深度学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]