Related Experiment Video
Updated: Jul 18, 2026
06:36
Three-Dimensional Mapping of the Rotation of Interactive Virtual Objects with Eye-Tracking Data
Published on: October 18, 2024
Intraclass retrieval of nonrigid 3D objects: application to face recognition
Georgios Passalis1, Ioannis A Kakadiaris, Theoharis Theoharis
1Department of Informatics and Telecommunications, University of Athens, Greece. passalis@di.uoa.gr
IEEE Transactions on Pattern Analysis and Machine Intelligence
|December 16, 2006
Summary
A new 3D object retrieval method uses class-specific models for accurate intraclass searching. This specialized approach enhances 3D database management and is particularly effective for tasks like face recognition.
Area of Science:
- Computer Science
- Computer Vision
- Pattern Recognition
Background:
- Managing large 3D object collections requires efficient retrieval methods.
- General retrieval techniques are insufficient for intraclass queries.
- Intraclass retrieval demands specialized methods leveraging class characteristics.
Purpose of the Study:
- To develop a novel 3D object retrieval method for intraclass tasks.
- To improve accuracy and efficiency in large-scale 3D databases.
- To address the challenges of nonrigid object retrieval.
Main Methods:
- Utilizing a parameterized annotated model incorporating class characteristics.
- Fitting the model using a deformable framework and converting to geometry images.
- Performing retrieval in the wavelet domain after transformation.
Main Results:
- Achieved high accuracy without user interaction.
- Demonstrated efficiency for large databases and nonrigid objects.
- Obtained a 95.2% verification rate in face recognition (FRGC v2 database).
Conclusions:
- The proposed method is effective for specialized intraclass 3D object retrieval.
- It offers a significant improvement over general retrieval methods.
- The approach is robust and applicable to challenging domains like face recognition.