Research

Text-based Search

Users can search for 3D models using keywords or descriptive text, similar to traditional text search engines.

2D Image Retrieval

Users can provide a 2D image as a query, and the system will return 3D models that are visually similar. This method often involves computer vision techniques such as image feature extraction and matching algorithms.

3D Shape Matching

Users can provide a known 3D shape as a query, and the system will search for similar models in the database. This typically involves shape descriptors and matching algorithms, such as 3D shape descriptors or deep learning models.

Semantic Search

This approach relies on semantic information about 3D models, allowing users to search using semantic labels, categories, or attributes. This requires annotating models with semantic labels or using automated annotation through deep learning models.

Content-based Retrieval

Content-based retrieval uses internal features or properties of models for searching, without relying on external text or image queries. Common methods include retrieval based on shape descriptors, color, texture, voxel grids, and more.

Content-based Retrieval

Users can refine search results by interacting with the system. For example, users can select a model and request the system to provide similar or dissimilar models, gradually refining the search results.

3D Model Retrieval Engines

Specialized software and search engines are designed for efficient 3D model retrieval, allowing users to search and filter models using various criteria.

Deep Learning-based Retrieval

In recent years, deep learning techniques have made significant advancements in 3D model retrieval. Neural network models can learn to extract useful features from model data, improving retrieval performance.