Hybrid associative retrieval of three-dimensional models

Area of Science:

• Computer Science
• Artificial Intelligence
• 3D Modeling

Background:

• Conventional 3-D model retrieval relies on similarity within the same class.
• Existing methods struggle with retrieving semantically related models from different categories.

Purpose of the Study:

• To propose a novel hybrid 3-D model associative retrieval framework.
• To enable retrieval of 3-D models that are semantically related but belong to different classes.

Main Methods:

• Developed three neural network-based approaches to establish semantic associations between 3-D model categories.
• Implemented a cyclic-shift scheme to partition associative groups into two-class pairwise groups.
• Utilized two strategies to combine retrieval results from different associative groups.

Main Results:

• Demonstrated the effectiveness of the proposed framework through experiments on various datasets.
• Showcased the efficiency of the hybrid associative retrieval approach.
• Successfully retrieved semantically related 3-D models across different classes.

Conclusions:

• The proposed hybrid 3-D model associative retrieval framework is effective and efficient.
• This novel approach expands the capabilities of 3-D model retrieval beyond simple similarity.
• The method successfully handles associative groups with varying numbers of classes.