[1]戴心杰,郑家杰,袁远飞,等.基于自监督与自适应感知关系网络的小样本图像分类[J].南京师范大学学报(工程技术版),2024,24(04):068-78.[doi:10.3969/j.issn.1672-1292.2024.04.007]
　Dai Xinjie,Zheng Jiajie,Yuan Yuanfei,et al.Few-Shot Image Classification Based on Self-Supervised and Adaptive-Aware Relation Network[J].Journal of Nanjing Normal University(Engineering and Technology),2024,24(04):068-78.[doi:10.3969/j.issn.1672-1292.2024.04.007]
点击复制

基于自监督与自适应感知关系网络的小样本图像分类
分享到：

南京师范大学学报（工程技术版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
24卷
期数:
2024年04期
页码:
068-78
栏目:
计算机科学与技术
出版日期:
2024-12-15

文章信息/Info

Title:
Few-Shot Image Classification Based on Self-Supervised and Adaptive-Aware Relation Network
文章编号:
1672-1292(2024)04-0068-11
作者:
戴心杰1; 2; 郑家杰1; 2; 袁远飞1; 2; 王李进1; 2; 吴清寿3
(1.福建农林大学计算机与信息学院,福建 福州 350002)
(2.福建农林大学智慧农林福建省高校重点实验室,福建 福州 350002)
(3.武夷学院数学与计算机学院,福建 武夷山 354300)
Author(s):
Dai Xinjie1; 2; Zheng Jiajie1; 2; Yuan Yuanfei1; 2; Wang Lijin1; 2; Wu Qingshou3
(1.College of Computer and Information Sciences,Fujian Agriculture and Forestry University,Fuzhou 350002,China)
(2.Key Laboratory of Smart Agricultrue and Foresty in Fujian Province University,Fujian Agriculture and Forestry University,Fuzhou 350002,China)
(3.School of Mathematics and Computer Science,Wuyi University,Wuyishan 354300,China)
关键词:
小样本分类; 自监督学习; 自适应感知关系网络; 度量学习; 双关联注意力机制; 动态权重平均
Keywords:
few-shot classification; self-supervised learning; adaptive-aware relation network; metric learning; dual correlated attention mechanism; dynamic weight averaging
分类号:
O643; X703
DOI:
10.3969/j.issn.1672-1292.2024.04.007
文献标志码:
A
摘要:
关系网络是通过度量分析样本之间相似性的小样本分类方法,其固有的局部连通性限制了对样本全局特征的利用,并且在数据量较少时,模型的泛化能力不足. 提出一种混合自监督学习和自适应感知关系网络的小样本分类方法. 首先,通过结合自监督的实例级和场景级辅助任务、有监督的小样本分类辅助任务和自适应双相关注意任务提升模型特征表示和泛化能力. 其次,引入动态权重平均策略,用于自适应优化辅助任务之间的权重. 实例级辅助任务用于学习旋转样本未知类别的转移知识,场景级辅助任务确保不同旋转数据集的分类器预测结果一致性,小样本分类辅助任务则对扩展数据集进行有监督的分类预测平均,优化分类效能. 自适应感知关系网络任务通过自适应层对图像特征变化进行自动调节,通过双关联注意力机制增强特征间相互作用,促进关键特征辨识. 在数据集miniImageNet、tieredImageNet和CUB-200-2011上进行了验证,提出的方法在不同的骨干网络上都能较好地提升关系网络的分类性能,表明该方法是可行有效的.
Abstract:
Relation networks,as a method for few-shot classification through metric analysis of sample similarities,are limited by their inherent local connectivity which restricts the utilization of global features of samples. Furthermore,these networks demonstrate insufficient generalization ability when data is scarce. This paper proposes a hybrid method of few-shot classification combining self-supervised learning with adaptive perception relation networks. Firstly,it enhances model feature representation and generalization ability by integrating self-supervised instance-level and scene-level auxiliary tasks,supervised few-shot classification auxiliary tasks,and adaptive dual-relation attention tasks. Additionally,a dynamic weight averaging strategy is introduced to adaptively optimize weights between auxiliary tasks. Instance-level auxiliary tasks focus on learning transfer knowledge of unknown categories in rotated samples,scene-level tasks ensure consistency in classifier predictions across different rotated datasets,while few-shot classification auxiliary tasks average supervised predictions on expanded datasets,optimizing classification efficacy. The adaptive perception relation network tasks automatically adjust image feature variations through an adaptive layer,and enhance inter-feature interactions via a dual-relation attention mechanism,thereby promoting key feature recognition. The proposed method has been validated on the miniImageNet,tieredImageNet and CUB-200-2011 datasets,demonstrating its capability to significantly enhance the classification performance of relation networks across various backbone networks,proving the feasibility and effectiveness of the proposed approach.

参考文献/References:

[1]LI F F,FERGUS R,PERONA P. One-shot learning of object categories[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2006,28(4):594-611.
[2]LAKE B M,SALAKHUTDINOV R R,TENENBAUM J. One-shot learning by inverting a compositional causal process[J]. Advances in Neural Information Processing Systems,2013,26:2526-2534.
[3]Yang J,LIU Y L. The latest advances in face recognition with single training sample[J]. Journal of Xihua University(Natural Science Edition),2014,33(4):1-5.
[4]KOTIA J,KOTWAL A,BHARTI R,et al. Few shot learning for medical imaging[J]. Machine Learning Algorithms for Industrial Applications,2021,907:107-132.
[5]CAI A H,HU W X,ZHENG J. Few-shot learning for medical image classification[C]//International Conference on Artificial Neural Networks. Bratislawa,Slovakia:Springer,2020:441-452.
[6]CHEN Z,EAVANI H,CHEN W,et al. Few-shot NLG with pre-trained language model[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg,PA:ACL,2020:183-190.
[7]HOWARD J,RUDER S. Universal language model fine-tuning for text classification[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics. Stroudsburg,PA:ACL,2018:328-339.
[8]RAVI S,LAROCHELLE H. Optimization as a model for few-shot learning[C]//International Conference on Learning Representations. Toulon,France,2017.
[9]FINN C,ABBEEL P,LEVINE S. Model-agnostic meta-learning for fast adaptation of deep networks[C]//International Conference on Machine Learning. Sydney,Australia,2017.
[10]LEE K,MAJI S,RAVICHANDRAN A,et al. Meta-learning with differentiable convex optimization[C]//CVF Conference on Computer Vision and Pattern Recognition. Long Beach,CA,USA,2019.
[11]LIU Y Y,SCHIELE B,SUN Q R. An ensemble of epoch-wise empirical bayes for few-shot learning[C]//Proceedings of the European Conference on Computer Vision. Glasgow,Scotland,UK:Springer,2020:404-421.
[12]VINYALS O,BLUNDELL C,LILLICRAP T,et al. Matching networks for one shot learning[C]//30th Conference on Neural Information Processing Systems. Barcelona,Spain,2016.
[13]SNELL J,SWERSKY K,ZEMEL R. Prototypical networks for few-shot learning[J]. 31th Conference on Neural Information Processing Systems. Long Beach,CA,USA,2017.
[14]SUNG F,YANG Y X,ZHANG L,et al. Learning to compare:Relation network for few-shot learning[C]//CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City,UT,USA,2018.
[15]ZHANG C,CAI Y J,LIN G S,et al. Deepemd:Few-shot image classification with differentiable earth mover's distance and structured classifiers[C]//CVF Conference on Computer Vision and Pattern Recognition. Virtual,2020.
[16]XIE J,LONG F,LV J,et al. Joint distribution matters:Deep brownian distance covariance for few-shot classification[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans,LA,USA,2022:7972-7981.
[17]HUI B,ZHU P,HU Q,et al. Self-attention relation network for few-shot learning[C]//Proceedings of the 2019 IEEE International Conference on Multimedia & Expo Workshops. Shanghai,China,2019:198-203.
[18]李晓旭,刘忠源,武继杰,等. 小样本图像分类的注意力全关系网络[ J]. 计算机学报,2023,46(2):371-384.
[19]WU Z,LI Y,GUO L,et al. PARN:Position-aware relation networks for few-shot learning[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Long Beach,CA,USA,2019:6659-6667.
[20]ABDELAZIZ M,ZHANG Z. Multi-scale kronecker-product relation networks for few-shot learning[J]. Multimedia Tools and Applications,2022,81(5):6703-6722.
[21]LI X,LI Y,ZHENG Y,et al. ReNAP:Relation network with adaptive prototypical learning for few-shot classification[J]. Neurocomputing,2023,520:356-364.
[22]HOU R,CHANG H,MA B,et al. Cross attention network for few-shot classification[J]. Advances in Neural Information Processing Systems,2019,32:4005-4006.
[23]LI Z,HU Z,LUO W,et al. SaberNet:Self-attention based effective relation network for few-shot learning[J]. Pattern Recognition,2023,133:109024.
[24]GIDARIS S,BURSUC A,KOMODAKIS N,et al. Boosting few-shot visual learning with self-supervision[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Long Beach,CA,USA,2019:8059-8068.
[25]ZHANG M,ZHANG J,LU Z,et al. IEPT:Instance-level and episode-level pretext tasks for few-shot learning[C]//International Conference on Learning Representations. Vienna,Austria,2021:1-16.
[26]GAO Y,FEI N,LIU G,et al. Contrastive prototype learning with augmented embeddings for few-shot learning[J]. Uncertainty in Artificial Intelligence,2021,21:140-150.
[27]YANG Z,WANG J,ZHU Y,et al. Few-shot classification with contrastive learning[C]//Proceedings of the European Conference on Computer Vision. Tel Aviv,Israel,2022:293-309.
[28]LIU S,JOHNS E,DAVISON A J,et al. End-to-end multi-task learning with attention[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach,CA,USA,2019:1871-1880.
[29]KINGMA D P,BA J. Adam:A method for stochastic optimization[C]//International Conference on Learning Representations. San Diego,CA,USA,2015:1-15.
[30]LAI J,YANG S,ZHOU J,et al. Clustered-patch element connection for few-shot learning[C]//International Joint Conference on Artificial Intelligence. San Francisco,CA,USA,2023:991-998.
[31]SUN Q,LIU Y,CHUA T S,et al. Meta-transfer learning for few-shot learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach,CA,USA,2019:403-412.
[32]QIN Z,WANG H,MAWULI C B,et al. Multi-instance attention network for few-shot learning[J]. Information Sciences,2022,611:464-475.
[33]YANG F,WANG R,CHEN X,et al. Semantic guided latent parts embedding for few-shot learning[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa,HI,USA,2023:5447-5457.
[34]SIMON C,KONIUSZ P,NOCK R,et al. Adaptive subspaces for few-shot learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle,WA,USA,2020:4136-4145.
[35]RAVICHANDRAN A,BHOTIKA R,SOATTO S,et al. Few-shot learning with embedded class models and shot-free meta training[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Long Beach,CA,USA,2019:331-339.
[36]YE H J,HU H,ZHAN D.C,et al. Few-shot learning via embedding adaptation with set-to-set functions[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle,WA,USA,2020:8808-8817.

备注/Memo

备注/Memo:
收稿日期:2024-05-12.
通讯作者:王李进,博士,教授,研究方向:智能计算. E-mail:lijinwang@fafu.edu.cn
更新日期/Last Update: 2024-12-15