面向元余弦损失的少样本图像分类  

Meta-cosine loss for few-shot image classification

作　　者：陶鹏 冯林[1] 杜彦东 龚勋[2] 王俊[3] Tao Peng;Feng Lin;Du Yandong;Gong Xun;Wang Jun(School of Computer Science,Sichuan Normal University,Chengdu 610101,China;School of Computing and Artificial Intelligence,Southwest Jiaotong University,Chengdu 610031,China;School of Business,Sichuan Normal University,Chengdu 610101,China)

机构地区：[1]四川师范大学计算机科学学院,成都610101 [2]西南交通大学计算机与人工智能学院,成都610031 [3]四川师范大学商学院,成都610101

出　　处：《中国图象图形学报》2024年第2期506-519,共14页Journal of Image and Graphics

基　　金：国家自然科学基金项目(61876158,71971151);四川省自然科学基金项目(2023NSFSC0210);四川省重点研发(2023YFS0202,2023YFG0267)。

摘　　要：目的 度量学习是少样本学习中一种简单且有效的方法,学习一个丰富、具有判别性和泛化性强的嵌入空间是度量学习方法实现优秀分类效果的关键。本文从样本自身的特征以及特征在嵌入空间中的分布出发,结合全局与局部数据增强实现了一种元余弦损失的少样本图像分类方法(a meta-cosine loss for few-shot image classification,AMCL-FSIC)。方法 首先,从数据自身特征出发,将全局与局部的数据增广方法结合起来,利于局部信息提供更具区别性和迁移性的信息,使训练模型更多关注图像的前景信息。同时,利用注意力机制结合全局与局部特征,以得到更丰富更具判别性的特征。其次,从样本特征在嵌入空间中的分布出发,提出一种元余弦损失(meta-cosine loss,MCL)函数,优化少样本图像分类模型。使用样本与类原型间相似性的差调整不同类的原型,扩大类间距,使模型测试新任务时类间距更加明显,提升模型的泛化能力。结果 分别在5个少样本经典数据集上进行了实验对比,在FC100(Few-shot Cifar100)和CUB(Caltech-UCSD Birds-200-2011)数据集上,本文方法均达到了目前最优分类效果;在MiniImageNet、TieredImageNet和Cifar100数据集上与对比模型的结果相当。同时,在MiniImageNet,CUB和Cifar100数据集上进行对比实验以验证MCL的有效性,结果证明提出的MCL提升了余弦分类器的分类效果。结论 本文方法能充分提取少样本图像分类任务中的图像特征,有效提升度量学习在少样本图像分类中的准确率。Objective Few-shot learning(FSL)is a popular and difficult problem in computer vision.It aims to achieveeffective classification with a few labeled samples.Recent few-shot learning methods can be divided into three major catego⁃ries:metric-,transfer-,and gradient-based methods.Among them,metric-based learning methods have received consider⁃able attention because of their simplicity and excellent performance in few-shot learning problems.In particular,metric-based learning methods consist of a feature extractor based on a convolutional neural network(CNN)and a classifier basedon spatial distance.By mapping the samples into the embedding space,a simple metric function is used to calculate thesimilarity between the sample and the class prototype,quickly identifying the novel class sample.The metric function isused for classification,and it bypasses the optimization problem in the few-shot setting when using network learning classifi⁃ers.Therefore,a richer,more discriminative,and better generalization embedding space is the key for metric-based learn⁃ing methods.From the perspective of the feature and its embedding space,and by combining the global and local featuresof a sample,we propose a meta-cosine loss for few-shot image classification method,called AMCL-FSIC,to improve theaccuracy of metric-based learning methods.Method On the one hand,our primary objective is to obtain suitable features.Image information is composed of foreground and background images.The foreground image is beneficial for few-shot classi⁃fication,whereas the background image is detrimental.If we can force the model to focus only on the foreground duringtraining and evaluation and disregard the background,then this scenario is helpful for image classification.However,it isnot easy to achieve.In fact,we need prior knowledge of the prospective object.As stated by previous researchers,imagesare roughly divided into global and local features,which are randomly cropped portions of each image.Local features con⁃tain cross-category discri

关 键 词：元学习 少样本学习(FSL) 度量学习 元余弦损失(MCL) 图像分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]