Abstract
The preference about poses of the idol varys and there is a demand for selection of idol images according to their pose. We present an agent for classifying by the pose of idol in still images. In this work, we focus on still images of idol wearing swimsuits. For each image, we create the feature vectors such as line segments indicating size and location, orientations of ten body parts(head, torso, upper/lower arms/legs) in reference to Eichner's Stickman Pose Estimation. Moreover, in order to raise the accuracy of the classification, we propose Human Pose Guide Ontology(HPGO) to guide and constrain the feature vectors. Finally, we evaluate our approach and show its efficiency of HPGO.
