SVM based ASM for facial landmarks location

Publisher:
IEEE
Publication Type:
Conference Proceeding
Citation:
8th IEEE International Conference on Computer and Information Technology, 2008. CIT 2008, 2008, pp. 321 - 326
Issue Date:
2008-01
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail2008000787OK.pdf1.06 MB
Adobe PDF
Finding a new position for each landmark is a crucial step in active shape model (ASM). Mahalanobis distance minimization is used for this finding, provided there are enough training data such that the grey-level profiles for each landmark follow a multivariate Gaussian distribution. However, this condition could not be satisfied in most cases. In this paper, a new method support vector machine (SVM) based ASM (SVMBASM) is proposed. It approaches the finding task as a small sample size classification problem, and uses SVM classifier to deal with this problem. Moreover, considering imbalanced dataset which contains more negative instances(incorrect candidates for new position) than positive instances (correct candidates for new position), a multi-class classification framework is adopted. Performance evaluation on SJTU face database show that the proposed SVMBASM outperforms the original ASM in terms of the average error as well as the average frequency of convergence.
Please use this identifier to cite or link to this item: