Face recognition with small and large size databases

TL;DR

This paper compares face recognition performance on small and large databases, specifically ORL with 40 subjects and FERET with 994, highlighting applications for security and large-scale identification.

Contribution

It provides experimental analysis of face recognition effectiveness across different database sizes, demonstrating practical implications for security and large user groups.

Findings

ORL database suitable for small-scale security applications

FERET database useful for large-scale user identification

Experimental results inform application-specific database choices

Abstract

This paper presents experimental results using the ORL (40 people) and FERET (994 people) databases. The ORL database can be useful for securing applications where few users attempting to access are expected. This is the case, for instance, of a PDA or PC where the password is the face of the user. On the other hand, the FERET database is useful for studying those situations where the number of authorized users is around a thousand people.