/face-recognition-framework-bio-inspired-algorithms

This repository contains the project from my master's thesis. It is well known that Face Recognition (FR) systems are still facing great challenges when variations of illumination, pose, expression and occlusion are present. Also, in many situations, it is only possible to obtain One Sample Per Person (OSPP) for training, representing a challenging real-world condition. In this work, we promote an FR solution that approaches the illumination variation challenge, along with the OSPP problem. The proposed FR framework is defined by an optimizer and a pool of preprocessing and feature extraction techniques. The approach makes available to the optimizer a pool of techniques, in which the optimizer seeks for the best set of strategies, also tuning their parameters. In this work, the FR framework uses the well-known Differential Evolution algorithm as the optimizer, called FR-DE. Two experimental methodologies are employed to assess the performance of the proposed FR-DE framework. The first methodology employs a standard dataset separation and uses the Yale Extended B dataset presenting severe illumination variation conditions. The second experimental methodology considers the OSPP problem along with illumination and poses variations. The CMU-PIE and FERET datasets are employed. The FR-DE is compared with some state-of-art algorithms and analysis suggest that the proposed framework is competitive and suitable for face recognition.

This repository is not active