Loading...
Please wait, while we are loading the content...
Similar Documents
Graph transduction learning with connectivity constraints with application to multiple foreground cosegmentation.
| Content Provider | CiteSeerX |
|---|---|
| Author | Ma, Tianyang Latecki, Longin Jan |
| Abstract | The proposed approach is based on standard graph transduction, semi-supervised learning (SSL) framework. Its key novelty is the integration of global connectivity constraints into this framework. Although connectivity leads to higher order constraints and their number is an exponential, finding the most violated connectivity constraint can be done efficiently in polynomial time. Moreover, each such constraint can be represented as a linear inequality. Based on this fact, we design a cutting-plane algorithm to solve the integrated problem. It iterates between solving a convex quadratic problem of label propagation with linear inequality constraints, and finding the most violated constraint. We demonstrate the benefits of the proposed approach on a realistic and very challenging problem of cosegmentation of multiple foreground objects in photo collections in which the foreground objects are not present in all photos. The obtained results not only demonstrate performance boost induced by the connectivity constraints, but also show a significant improvement over the state-of-the-art methods. apple +bucket baby girl+blue girl+red Figure 1. Multiple Foreground Cosegmentation results on three images of the scene Apple+picking. First Columns: original images. Second Columns: the results of an excellent graph transduction SSL method RLGC [24]. Third Column: results of the proposed GTC. Compared to RLGC, GTC improves the consistency of label assignment by enforcing connectivity of regions with the same label. 1. |
| File Format | |
| Access Restriction | Open |
| Subject Keyword | Connectivity Constraint Multiple Foreground Cosegmentation Graph Transduction Learning Semi-supervised Learning Second Column Demonstrate Performance Boost Standard Graph Transduction Violated Connectivity Constraint Foreground Object Obtained Result Label Propagation Linear Inequality Constraint Order Constraint Cutting-plane Algorithm Violated Constraint Label Assignment Photo Collection Key Novelty State-of-the-art Method Challenging Problem Linear Inequality Scene Apple Picking Third Column Polynomial Time Global Connectivity Constraint Multiple Foreground Object Original Image Convex Quadratic Problem Multiple Foreground Cosegmentation Result Significant Improvement First Column |
| Content Type | Text |
