The nature of statistical learning theory
The nature of statistical learning theory
An introduction to support Vector Machines: and other kernel-based learning methods
An introduction to support Vector Machines: and other kernel-based learning methods
Detecting Faces in Images: A Survey
IEEE Transactions on Pattern Analysis and Machine Intelligence
A decision-theoretic generalization of on-line learning and an application to boosting
EuroCOLT '95 Proceedings of the Second European Conference on Computational Learning Theory
Face recognition: A literature survey
ACM Computing Surveys (CSUR)
Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
Automated annotation of human faces in family albums
MULTIMEDIA '03 Proceedings of the eleventh ACM international conference on Multimedia
Fast learning in networks of locally-tuned processing units
Neural Computation
The Digital Album: A Personal File-tainment System
ICMCS '96 Proceedings of the 1996 International Conference on Multimedia Computing and Systems
Fast rotation invariant multi-view face detection based on real adaboost
FGR' 04 Proceedings of the Sixth IEEE international conference on Automatic face and gesture recognition
Hi-index | 0.00 |
In this paper, we propose a novel approach to automatically index digital home photos based on person identity. A person is identified by his/her face and clothes. The proposed method consists of two parts: clustering and indexing. In the clustering, a series of unlabeled photos is aligned in taken-time order, and is divided into several sub-groups by situation. The situation groups are decided by time and visual differences. In the indexing, SVMs are trained with features of pre-indexed faces to model target persons. The representative feature vector of the person group from the clustering is queried to the trained SVMs. Each SVM outputs a numeric confidence value about the query person group. The query person group is determined to the target person by the most confident SVM. The experimental results showed that the proposed method outperformed traditional person indexing method using only face feature and its performance increased to 93.56% from 72.31%.