Mapping features to semantic space for 3D model retrieval

In this paper, a novel offline supervised learning method is proposed to map low-level features to high-level semantic space for 3D model retrieval. Firstly, users are expected to retrieval models through relevance feedback learning, and the retrieval history will be recorded in a distance matrix which can be renewed gradually. Then, Laplacian Eigemnaps is used to establish the representations of models in semantic space by means of retrieval history. Experimental evaluation on the standard repositories PSB shows that our method can effectively capture the user's knowledge in mind and embody it in the retrieval results. The algorithm also displays its strong generalization ability in our experiments.