TY - GEN
T1 - Unsupervised visual representation learning by graph-based consistent constraints
AU - Li, Dong
AU - Hung, Wei Chih
AU - Huang, Jia Bin
AU - Wang, Shengjin
AU - Ahuja, Narendra
AU - Yang, Ming Hsuan
N1 - Publisher Copyright:
© Springer International Publishing AG 2016.
PY - 2016
Y1 - 2016
N2 - Learning rich visual representations often require training on datasets of millions of manually annotated examples. This substantially limits the scalability of learning effective representations as labeled data is expensive or scarce. In this paper, we address the problem of unsupervised visual representation learning from a large, unlabeled collection of images. By representing each image as a node and each nearest-neighbor matching pair as an edge, our key idea is to leverage graph-based analysis to discover positive and negative image pairs (i.e., pairs belonging to the same and different visual categories). Specifically, we propose to use a cycle consistency criterion for mining positive pairs and geodesic distance in the graph for hard negative mining. We show that the mined positive and negative image pairs can provide accurate supervisory signals for learning effective representations using Convolutional Neural Networks (CNNs). We demonstrate the effectiveness of the proposed unsupervised constraint mining method in two settings: (1) unsupervised feature learning and (2) semi-supervised learning. For unsupervised feature learning, we obtain competitive performance with several state-of-the-art approaches on the PASCAL VOC 2007 dataset. For semisupervised learning, we show boosted performance by incorporating the mined constraints on three image classification datasets.
AB - Learning rich visual representations often require training on datasets of millions of manually annotated examples. This substantially limits the scalability of learning effective representations as labeled data is expensive or scarce. In this paper, we address the problem of unsupervised visual representation learning from a large, unlabeled collection of images. By representing each image as a node and each nearest-neighbor matching pair as an edge, our key idea is to leverage graph-based analysis to discover positive and negative image pairs (i.e., pairs belonging to the same and different visual categories). Specifically, we propose to use a cycle consistency criterion for mining positive pairs and geodesic distance in the graph for hard negative mining. We show that the mined positive and negative image pairs can provide accurate supervisory signals for learning effective representations using Convolutional Neural Networks (CNNs). We demonstrate the effectiveness of the proposed unsupervised constraint mining method in two settings: (1) unsupervised feature learning and (2) semi-supervised learning. For unsupervised feature learning, we obtain competitive performance with several state-of-the-art approaches on the PASCAL VOC 2007 dataset. For semisupervised learning, we show boosted performance by incorporating the mined constraints on three image classification datasets.
KW - Convolutional neural networks
KW - Image classification
KW - Semi-supervised learning
KW - Unsupervised feature learning
UR - http://www.scopus.com/inward/record.url?scp=84990065848&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84990065848&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-46493-0_41
DO - 10.1007/978-3-319-46493-0_41
M3 - Conference contribution
AN - SCOPUS:84990065848
SN - 9783319464923
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 678
EP - 694
BT - Computer Vision - 14th European Conference, ECCV 2016, Proceedings
A2 - Leibe, Bastian
A2 - Matas, Jiri
A2 - Sebe, Nicu
A2 - Welling, Max
PB - Springer
T2 - 14th European Conference on Computer Vision, ECCV 2016
Y2 - 11 October 2016 through 14 October 2016
ER -