Learning High-Level Feature by Deep Belief Networks for 3-D Model Retrieval and Recognition

TLDR

3‑D shape analysis has attracted extensive research efforts, with the major challenge being the design of effective high‑level 3‑D shape features. This paper proposes a multi‑level 3‑D shape feature extraction framework using deep learning. Low‑level 3‑D descriptors are encoded into a geometric bag‑of‑words, from which middle‑level patterns are discovered, and high‑level features are then learned via deep belief networks to capture discriminative geometric relationships. Experiments on shape recognition and retrieval show that the proposed method outperforms state‑of‑the‑art approaches.

Abstract

3-D shape analysis has attracted extensive research efforts in recent years, where the major challenge lies in designing an effective high-level 3-D shape feature. In this paper, we propose a multi-level 3-D shape feature extraction framework by using deep learning. The low-level 3-D shape descriptors are first encoded into geometric bag-of-words, from which middle-level patterns are discovered to explore geometric relationships among words. After that, high-level shape features are learned via deep belief networks, which are more discriminative for the tasks of shape classification and retrieval. Experiments on 3-D shape recognition and retrieval demonstrate the superior performance of the proposed method in comparison to the state-of-the-art methods.