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.
|Title of host publication||Computer Vision - 14th European Conference, ECCV 2016, Proceedings|
|Editors||Bastian Leibe, Jiri Matas, Nicu Sebe, Max Welling|
|Number of pages||17|
|Publication status||Published - 2016|
|Event||14th European Conference on Computer Vision, ECCV 2016 - Amsterdam, Netherlands|
Duration: 2016 Oct 8 → 2016 Oct 16
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||14th European Conference on Computer Vision, ECCV 2016|
|Period||16/10/8 → 16/10/16|
Bibliographical noteFunding Information:
This work is supported in part by the Initiative Scientific Research Program of Ministry of Education under Grant #20141081253. J.-B. Huang and N. Ahuja are supported in part by Office of Naval Research under Grant N00014-16-1-2314. W.-C. Hung and M.-H. Yang are supported in part by the NSF CAREER Grant #1149783 and gifts from Adobe and Nvidia.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)