Multi-Level Recommendation for Sketch Based 3D Toy Modeling

Abstract

Models recommendation is always a key part in part-composition system based on sketch. In this paper, proper models are suggested via three steps. Firstly, the possible type of model parts is suggested according to the topological features. Then toy parts that have similar profile types with the inputted sketches are clustered and recommended by lineage clustering method. Finally, based on the promoted $1 recognizer, the suggested parts are sorted according to the shape similarity with the inputted sketch. Experimental results show that the models can be recommended quickly and properly via our method.