Harnessing noisy Web images for deep representation

•A novel approach that uses convnets to learn transferable image representation based on a massive amount of (noisy) Web images.•Image reranking algorithms are useful to improve the generalization of convnet by refining the noisy training database.•Deeper convnet architectures can be trained on nois...

Full description

Saved in:
Bibliographic Details
Published in:Computer vision and image understanding 2017-11, Vol.164, p.68-81
Main Authors: Vo, Phong D., Ginsca, Alexandru, Le Borgne, Hervé, Popescu, Adrian
Format: Article
Language:English
Subjects:
Computer Science
Convolutional networks
Deep learning
Domain adaptation
Engineering Sciences
Noisy data
Representation learning
Semi-supervised learning
Signal and Image processing
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•A novel approach that uses convnets to learn transferable image representation based on a massive amount of (noisy) Web images.•Image reranking algorithms are useful to improve the generalization of convnet by refining the noisy training database.•Deeper convnet architectures can be trained on noisy images in order to alleviate for low (noisy) quality of the data. The keep-growing content of Web images is probably the next important data source to scale up deep neural networks which recently surpass human in image classification tasks. The fact that deep networks are hungry for labelled data limits themselves from extracting valuable information of Web images which are abundant and cheap. There have been efforts to train neural networks such as autoencoders with respect to either unsupervised or semi-supervised settings. Nonetheless they are less performant than supervised methods partly because the loss function used in unsupervised methods, for instance Euclidean loss, failed to guide the network to learn discriminative features and ignore unnecessary details. We instead train convolutional networks in a supervised setting but use weakly labelled data which are large amounts of unannotated Web images downloaded from Flickr and Bing. Our experiments are conducted at several data scales, with different choices of network architecture, and alternating between different data preprocessing techniques. The effectiveness of our approach is shown by the good generalization of the learned representations with new six public datasets.
ISSN:1077-3142
1090-235X
DOI:10.1016/j.cviu.2017.01.009