Homotopic image pseudo-invariants for openset object recognition and image retrieval

Area of Science:

• Computer Vision
• Biometrics
• Image Processing

Background:

• Face recognition is crucial for security and identification.
• Existing methods face challenges with open-set recognition and large datasets.
• Pixelwise analysis offers a potential avenue for improved recognition accuracy.

Purpose of the Study:

• To develop novel homotopic image pseudo-invariants for robust face recognition.
• To enable effective open-set face recognition using pixelwise analysis.
• To facilitate internet-scale recognition of faces and other objects.

Main Methods:

• Pixelwise analysis to compute homotopic image pseudo-invariants.
• Initial matching of exemplar and test images to find the most similar.
• Utilizing pseudo-invariants to verify identity, enabling open-set recognition.

Main Results:

• The proposed method demonstrates effectiveness in face recognition tasks.
• Recognition performance is enhanced when utilizing a face database.
• The approach is validated on the FERET dataset and internet-downloaded images.

Conclusions:

• Homotopic image pseudo-invariants offer a promising approach for face recognition.
• The method is applicable to open-set recognition and internet-based object identification.
• The technique shows potential for real-world applications requiring high recognition rates.