A semi-supervised network based on feature embeddings for image classification

被引:2
作者
Nuhoho, Raphael Elimeli [1 ]
Chen Wenyu [1 ]
Baffour, Adu Asare [2 ]
机构
[1] Univ Elect Sci & Technol China, Computat Intelligence Lab, Sch Comp Sci & Technol, 2006 Xiyuan Ave, Chengdu 611731, Sichuan, Peoples R China
[2] Univ Elect Sci & Technol China, Sch Informat & Software Engn, Chengdu, Peoples R China
关键词
clustering; convolutional neural network; feature embedding; semi-supervised learning;
D O I
10.1111/exsy.12908
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Deep learning approaches, including convolutional neural networks, are suitable for image classification tasks with well-labelled data. Unfortunately, we do not always have sufficiently labelled data. Recent methods attempt to leverage labelled and unlabelled data using fine-tuning or transfer learning. However, these methods rely on low-level image features. This article departs from recent works and proposes a new semi-supervised learning network that constitutes a convolutional branch and a neighbour cluster branch. Also, we introduce a new loss function that carefully optimizes the network according to the labelled/unlabelled data. In this way, we reduce any tendency to rely on low-level features, which is the case in current methods. We use datasets from three different domains (hand-written digits, natural images, and objects) to analyse the performance of our method. Experimental analysis shows that the network performs better by learning inherent discrimination features when integrating unlabelled data into the model's training process. Our proposed approach also provides strong generalization in the context of transfer learning. Finally, this study shows that the proposed loss function optimizes the network to produce more efficient feature embeddings for domain adaptation.
引用
收藏
页数:16
相关论文
共 35 条
[1]  
[Anonymous], 2015, ACS SYM SER
[2]  
[Anonymous], 2010, P ICML 2010 P 27 INT
[3]   Spatial self-attention network with self-attention distillation for fine-grained image recognitionx2729; [J].
Baffour, Adu Asare ;
Qin, Zhen ;
Wang, Yong ;
Qin, Zhiguang ;
Choo, Kim-Kwang Raymond .
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION, 2021, 81
[4]   Deep Clustering for Unsupervised Learning of Visual Features [J].
Caron, Mathilde ;
Bojanowski, Piotr ;
Joulin, Armand ;
Douze, Matthijs .
COMPUTER VISION - ECCV 2018, PT XIV, 2018, 11218 :139-156
[5]  
Coates Adam., 2011, Proceedings of Machine Learning Research, P215
[6]  
Dai AM, 2015, ADV NEUR IN, V28
[7]  
de Mello R. F., 2018, Machine Learning, DOI [DOI 10.1007/978-3-319-94989-5_4, DOI 10.1007/978-3-319-94989-5]
[8]   Robust feature spaces from pre-trained deep network layers for skin lesion classification [J].
dos Santos, Fernando Pereira ;
Ponti, Moacir A. .
PROCEEDINGS 2018 31ST SIBGRAPI CONFERENCE ON GRAPHICS, PATTERNS AND IMAGES (SIBGRAPI), 2018, :189-196
[9]   More discriminative convolutional neural network with inter-class constraint for classification [J].
Fei, Jianchao ;
Rui, Ting ;
Song, Xiaona ;
Zhou, You ;
Zhang, Sai .
COMPUTERS & ELECTRICAL ENGINEERING, 2018, 68 :484-489
[10]   A semi-supervised convolutional neural network based on subspace representation for image classification [J].
Gatto, Bernardo B. ;
Souza, Lincon S. ;
dos Santos, Eulanda M. ;
Fukui, Kazuhiro ;
Junior, Waldir S. S. ;
dos Santos, Kenny V. .
EURASIP JOURNAL ON IMAGE AND VIDEO PROCESSING, 2020, 2020 (01)