Learning image features with fewer labels using a semi-supervised deep convolutional network

Autor(es)

dos Santos, Fernando P. ^[1] ; Zor, Cemre ^[2] ; Kittler, Josef ^[3] ; Ponti, Moacir A. ^[1]

Afiliação do(s) autor(es)

^[1] Univ Sao Paulo, Inst Math & Comp Sci ICMC, BR-13566590 Sao Carlos, SP - Brazil

^[2] UCL, Ctr Med Image Comp CMIC, London WC1E 7JE - England

^[3] Univ Surrey, Ctr Vis Speech & Signal Proc CVSSP, Guildford GU2 7XH, Surrey - England

• Tipo de documento: Artigo Científico
• Fonte: NEURAL NETWORKS; v. 132, p. 131-143, DEC 2020.
• Citações Web of Science: 0

Resumo

Learning feature embeddings for pattern recognition is a relevant task for many applications. Deep learning methods such as convolutional neural networks can be employed for this assignment with different training strategies: leveraging pre-trained models as baselines; training from scratch with the target dataset; or fine-tuning from the pre-trained model. Although there are separate systems used for learning features from labelled and unlabelled data, there are few models combining all available information. Therefore, in this paper, we present a novel semi-supervised deep network training strategy that comprises a convolutional network and an autoencoder using a joint classification and reconstruction loss function. We show our network improves the learned feature embedding when including the unlabelled data in the training process. The results using the feature embedding obtained by our network achieve better classification accuracy when compared with competing methods, as well as offering good generalisation in the context of transfer learning. Furthermore, the proposed network ensemble and loss function is highly extensible and applicable in many recognition tasks. (c) 2020 Elsevier Ltd. All rights reserved. (AU)

• Processo FAPESP: 19/07316-0 - Teoria de singularidades e aplicações a geometria diferencial, equações diferenciais e visão computacional
• Beneficiário: Farid Tari
• Modalidade de apoio: Auxílio à Pesquisa - Temático
• Processo FAPESP: 18/22482-0 - Aprendendo características de conteúdo visual sob condições de supervisão limitada utilizando múltiplos domínios
• Beneficiário: Moacir Antonelli Ponti
• Modalidade de apoio: Auxílio à Pesquisa - Regular