Abstract
Face recognition using a single reference image per subject is challenging, above all when referring to a large gallery of subjects. Furthermore, the problem hardness seriously increases when the images are acquired in unconstrained conditions. In this paper we address the challenging Single Sample Per Person (SSPP) problem considering large datasets of images acquired in the wild, thus possibly featuring illumination, pose, face expression, partial occlusions, and low-resolution hurdles. The proposed technique alternates a sparse dictionary learning technique based on the method of optimal direction and the iterative -norm minimization algorithm called k-LiMapS. It works on robust deep-learned features, provided that the image variability is extended by standard augmentation techniques. Experiments show the effectiveness of our method against the hardness introduced above: first, we report extensive experiments on the unconstrained LFW dataset when referring to large galleries up to 1680 subjects; second, we present experiments on very low-resolution test images up to pixels; third, tests on the AR dataset are analyzed against specific disguises such as partial occlusions, facial expressions, and illumination problems. In all the three scenarios our method outperforms the state-of-the-art approaches adopting similar configurations.
Highlights
Faces convey a plethora of information, such as expression, gender, age, ethnic origin, and identity. these factors co-exist and the ability to recognize each of them is strictly correlated with the capability to isolate one from the others
This paper presents a method, called Sparsity-driven Sub-dictionary Learning using Deep features (SSLD), for solving the Single Sample Per Person (SSPP) problem coupled with other hurdles which arise from large-scale datasets, large appearance variations, and LR probe images
The proposed technique consists in a sparse-driven sub-dictionary learning strategy exploiting the richness of the augmented face image step, the strength of deep features, the simplicity of the method of optimal directions (MOD) technique for sub-dictionary learning, and the effectiveness of the sparse representation via k-L I M AP S on structured dictionaries
Summary
Faces convey a plethora of information, such as expression, gender, age, ethnic origin, and identity. The feature space is obtained employing deep features coupled with the linear discriminant analysis, while the concise model is derived adopting the method of optimal directions (MOD) [13], which has proved to be very efficient for low-dimensional input data The benefits of this approach is that, contrarily to generic learning algorithms [14], the label consistency between dictionary atoms and training data is maintained, allowing the direct application of the classification stage based on majority voting (a demo code is available on the website: https://github.com/phuselab/SSLD-face_recognition). The core idea in our Sparsity-driven Sub-dictionary Learning using Deep features (SSLD) technique is to work out a large number of face augmentation, characterize them with very discriminative deep features, derive a succinct sub-dictionary for each subject through k-L I M AP S sparse optimizer, and deduce the identity of probe images by combining multiple classifications by the majority voting This pipeline allows to deal with SSPP problem coupled with several further nuisances, while keeping the system very efficient, and suitable for real-world applications.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
