Pose Estimation Based on Semi-Supervised Latent Dirichlet Allocation

WEI Wei,ZHANG Yan-ning
2011-01-01
Abstract:Topics cannot be labeled in the unsupervised topic model,while the labeling work in supervised topic models is tedious and subjective.To solve these problems,a semi-supervised topic model was proposed.First,the location-irrelevant local features were detected and described by the scale-invariant feature transform(SIFT),based on which images were represented by a bag of visual words.Then partial labels were introduced to the topic-word level distribution in the latent Dirichlet allocation(LDA) model to guide the classification of the unlabeled data,which resulted in a semi-supervised LDA(SSLDA) model.The validation on head pose estimation showed the classification rate of the proposed method was 9.0%~24.7% higher than that of LDA.And the pose classification rate on partially occluded and misaligned face images was 8.8% and 21.5%~39.8% higher than multi-pose PCA method.With a small amount of labeled images,the proposed SSLDA model approaches the fully supervised LDA method.And it is applicable to other image classification problems.
What problem does this paper attempt to address?