A Learning-Based Approach For Perceptual Models Of Preference

This paper introduces a novel data-driven approach based on subjective constraints and feature learning for training perceptual models of preference. Fuzzy evaluation is applied to describe the subjective opinions from a large set of data collected from user study. Combined with the objective attributes of the training models and the subjective preferences, an optimization method is developed successfully for training and learning perceptual models. Two applications are given in details for the selection of "best" viewpoint of 3D objects and the optimized direction of 3D printing, which verify the effectiveness of our approach. This work also demonstrate a good human-computer interaction practice that draws supporting knowledge from both the machine side and the human side.