Unsupervised 3D Object Retrieval in Loop View

For 3D object retrieval,traditional multi-view methods usually rely on supervised classification for feature learning,which requires lots of manual efforts to annotate data.Differently,we present an unsupervised scheme to deal with the problem by using loop view data.First,we perform unsupervised deep learning of 3D objects for multiple loop view data and we leverage random data mixture to learn the latent relations between different shapes.Then,we introduce an optimal matching method for similarity matching,where the optimal value is calculated by averaging the minimized loop view distances.Finally,we propose a filtering algorithm for loop view features to reduce data redundancy,which can significantly save the computational cost yet preserving the retrieval accuracy.We carry out experiments on two public datasets ModelNet40 and SHREC15.The experimental results show that,compared with related methods,our algorithm has achieved excellent performance for unsupervised 3D object retrieval without requiring data annotation.