Images Classification with Limited Number of Labeled Data Using Domain Adaptation

Please enable javascript in your browser.

Taheri, Sahar | 2015

609 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 47554 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Jamzad, Mansour
Abstract:
The traditional machine learning methods assume that the training data and the test data are drawn from the same distribution (or drawn from the same domain). In practice, in many computer vision applications, this assumption may not hold. Unfortunately, the performance of these methods degrades on dataset drawn from a different domain. Domain adaptation attempts to minimize this degradation caused by distribution mismatch between the training and test data. Domain adaptation tries to adapt a model trainded from one domain to another domain. We focus on supervised domain adaptation method in which limited labeled data is available from the target domain. We propose a new domain adaptation algorithm called Cluster-based Projective Model Transfer SVM (CPMT-SVM) for images classification. This algorithm can be extended to multiple source domain adaptation problems. Our experiments show that the proposed approach consistently achieves the state-of-the-art on the standard Office-Caltech benchmark dataset for domain adaptation
Keywords:
Domain Adaptation ; Images Classification ; Target Domain ; Multiple Source Domain Adaptation

No TOC