Face images captured in unconstrained environments usually contain pose variation, which significantly degrades the performance of algorithms designed to recognize frontal faces.
This project implements a face identification framework capable of handling pose variations within ±45° of yaw. The project used the following paper as referance for methods and approaches
A Comprehensive Survey on Pose-Invariant Face Recognition
https://arxiv.org/abs/1502.04383Abstract: The capacity to recognize faces under varied poses is a fundamental human ability that presents a unique challenge for computer vision systems. Compared to frontal face recognition, which has been intensively studied and has gradually matured in the past few decades, pose-invariant face recognition (PIFR) remains a largely unsolved problem. However, PIFR is crucial to realizing the full potential of face recognition for real-world applications, since face recognition is intrinsically a passive biometric technology for recognizing uncooperative subjects. In this paper, we discuss the inherent difficulties in PIFR and present a comprehensive review of established techniques. Existing PIFR methods can be grouped into four categories, i.e., pose-robust feature extraction approaches, multi-view subspace learning approaches, face synthesis approaches, and hybrid approaches. The motivations, strategies, pros/cons, and performance of representative appraches are described and compared. Moreover, promising directions for future research are discussed.
The proposed framework first transforms the original pose-invariant face recognition problem into a partial frontal face recognition problem. A robust patch-based face representation scheme is then developed to represent the synthesized partial frontal faces.
For each patch, a transformation dictionary is learnt under the proposed multitask learning scheme. The transformation dictionary transforms the features of different poses into a discriminative subspace.
Finally, face matching is performed using a PCA-based Face Recognition system called ‘Eigenface’ where the synthesised frontal face is matched with the correct frontal face from within a frontal face dataset. Experimentation was done using captured image of a subject at multiple angles and the frontalization algorithm was tested on multiple images in the LFW (Labelled Faces in the Wild) dataset.
Details on running the code for frontalization and face-matching and all citations are in the respective readme files
Further details can be found in the report and in other relevant material