Impact Analysis of Different Effective Loss Functions by Using Deep Convolutional Neural Network for Face Recognition

Deep learning, in particular Convolutional Neural Network (CNN), has achieved promising results in face recognition recently. However, it remains an open question: why CNNs work well and how to design a 'good' architecture. The existing works tend to focus on reporting CNN architectures that work well for face recognition rather than investigate the reason. In this work, we conduct an extensive evaluation of CNN-based face recognition systems (CNN-FRS) on a common ground to make our work easily reproducible. Specifically, we use public database LFW (Labeled Faces in the Wild) to train CNNs, unlike most existing CNNs trained on private databases. We propose three CNN architectures which are the first reported architectures trained using LFW data. This paper quantitatively compares the architectures of CNNs and evaluate the effect of different implementation choices. We identify several useful properties of CNN-FRS. For instance, the dimensionality of the learned features can be significantly reduced without adverse effect on face recognition accuracy. In addition, traditional metric learning method exploiting CNN-learned features is evaluated. Experiments show two crucial factors to good CNN-FRS performance are the fusion of multiple CNNs and metric learning. To make our work reproducible, source code and models will be made publicly available.

Illumination, pose variations, disguises, aging effects and expression variations are some of the key factors that affect the performance of face recognition systems. Face recognition systems have always been studied from a recognition perspective. Our emphasis is on deriving a measure of similarity between faces. The similarity measure provides insights into the role each of the above mentioned variations play in affecting the performance of face recognition systems. In the process of computing the similarity measure between faces, we suggest a framework to compensate for pose variations and introduce the notion of 'half-faces' to circumvent the problem of non-uniform illumination. We used the similarity measure to retrieve similar faces from a database containing multiple images of individuals. Moreover, we devised experiments to study the effect age plays in affecting facial similarity. In conclusion, the similarity measure helps in studying the significance facial features play in affecting the performance of face recognition systems.

Face aging simulation and recognition based on NMF algorithm with sparseness constraints

The performance of face recognition systems can be dramatically degraded when the pose of the probe face is different from the gallery face. In this paper, we present a pose robust face recognition model, centered on modeling how face patches change in appearance as the viewpoint varies. We present a novel model based on two robust local appearance descriptors, Gabor wavelets and Local Binary Patterns (LBP). These two descriptors have been widely exploited for face recognition and different strategies for combining them have been investigated. However, to the best of our knowledge, all existing combination methods are designed for frontal face recognition. We introduce a local statistical framework for face recognition across pose variations, given only one frontal reference image. The method is evaluated on the Feret pose dataset and experimental results show that we achieve very high recognition rates over the wide range of pose variations presented in this challenging dataset.

Hybrid meta-heuristic algorithm based deep neural network for face recognition

Prior research has shown that the performance of face recognition systems in variant expression condition degrade seriously compared with invariant expression condition. Face recognition with variant expression is a challenging problem. In this paper, SIFT method is proposed to research on face recognition with variant expression. Two experiments using SIFT method are performed on a variant expression face database. In experiment 1, two images of one person with different expression are compared. In experiment 2, two images of different persons with the same expression are compared. Experimental results show that SIFT method could overcome the whole comparability of different faces and extract the local detail features of face. The experiment demostrates the huge potential of SIFT method in application to face recognition with variant expression.

At present, the performance of face recognition system depends much on the variations in illumination. To solve this problem, this paper presents an adaptable face recognition approach that uses filter fusion representation. The key idea is to use context-aware filter fusion to get better image from a bad illumination one. Genetic algorithm is the tool for adaptation for individual context category. These can provide robust face recognition on illumination context-awareness under uneven environments. Gabor wavelet representation can also provide a robust feature for image enhancement. Using these approaches, we have developed a robust face recognition technique that can recognize with a notable success and it has been tested on Inha DB and FERET face images.

Face recognition at a distance is one grand challenge for security surveillance. In this paper, the face images at different distances are simulated by varying image scales (resolutions). The performances of three face recognition algorithms (matchers) are tested with variant image scales (simulating different distances) and with two spectral images (modalities). The three selected matchers are face pattern byte, elastic bunch graph matching, and linear discriminant analysis; while the two modalities are visible and thermal images. The performance of a face recognition system can be measured by accuracy (AC) rate and false accept rate (FAR). To enhance the performance of face recognition especially at a distance, score fusion techniques are applied, which combine several scores from multiple matchers and multiple modalities. Our experiments are conducted with the ASUMS face dataset consisting of two spectral images (visible and thermal) from 135 subjects. The experimental results show that the face recognition with small image scales (simulating long distances) have low performance (e.g., AC=91.36%, FAR=8.64% for 20×20-pixel images); and score fusion can greatly improve accuracy (99.34%) meanwhile reduce FAR (0.31%).

An empirical study of the impact of masks on face recognition

Pose variation which brings illumination change, occlusion and non-linear scale variation, dramatically drops the performance of face recognition systems. In this paper, we propose a novel pose invariant face recognition method, in which we build a joint sparse coding scheme to predict face images from a certain pose to another. By introducing autoregressive regularization and symmetric information, our algorithm could achieve high robustness to local misalignment and large pose differences. Besides, we propose a new coarse pose estimation algorithm by collaborative representation classifier, which is very fast and enough accurate for our synthesis algorithm. The experiment results on multi-pose subsets of CMU-PIE and FERET database show the efficiency of the proposed method on multi-pose face recognition.

In recent years much advancement have been made in face recognition techniques to cater to the challenges such as pose, expression, illumination, aging and disguise. However, due to advances in technology, there are new emerging challenges for which the performance of face recognition systems degrades and plastic/cosmetic surgery is one of them. In this paper we comment on the effect of plastic surgery face image in multimodal biometric face recognition using speech signal. Speaker identity is correlated with the physiological and behavioral characteristics of the speaker. Selecting the most effective fusion techniques depends on operational issues such as accuracy requirements, availability of training data, and the validity of simplifying assumptions.

This paper investigates the possibility of exploiting facial skin texture regions to further improve the performance of face recognition systems. Information extracted from the forehead region is combined with scores produced by a kernel-based face recognition algorithm in a novel framework that can adapt to the availability of pure skin patches. A novel skin/non-skin classifier is presented for detecting such pure skin patches in the forehead region using state-of-the-art texture feature extraction techniques. The pure-skin forehead image regions are then classified using a sparse representation classifier to produce scores which are fused with the results of whole-face classifiers. The proposed algorithm is tested using the XM2VTS database and compared with other results published using similar protocols. The results suggest that exploiting pure skin regions in such an adaptive framework could significantly enhance recognition accuracy.KeywordsFace RecognitionFace ImageSparse RepresentationSkin RegionIdentity InformationThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

In this study, the face recognition task is applied on masked and unmasked faces using hand-crafted methods. Due to COVID-19 and masks, facial identification from unconstrained images became a hot topic. To avoid COVID-19, most people use masks outside. In many cases, typical facial recognition technology is useless. The majority of contemporary advanced face recognition methods are based on deep learning, which primarily relies on a huge number of training examples, however, masked face recognition may be investigated using hand-crafted approaches at a lower computing cost than using deep learning systems. A low-cost system is intended to be constructed for recognizing masked faces and compares its performance to that of face recognition systems that do not use masks. The proposed method fuses hand-crafted methods using feature-level fusion strategy. This study compares the performance of masked and unmasked face recognition systems. The experiments are undertaken on two publicly accessible datasets for masked face recognition: Masked Labeled Faces in the Wild (MLFW) and Cross-Age Labeled Faces in the Wild (CALFW). The best accuracy is achieved as 94.8% on MLFW dataset. The rest of the results on different train and test sets from CALFW and MLFW datasets are encouraging compared to the state-of-the-art models.

Face images obtained by an outdoor surveillance camera, are often confronted with severe degradations (e.g., low-resolution, low-contrast, blur and noise). This significantly limits the performance of face recognition (FR) systems. This paper presents a framework to overcome the degradation in images obtained by an outdoor surveillance camera, to improve the performance of FR. We have defined a measure that is based on the difference in intensity histograms of face images, to estimate the amount of degradation. In the past, super-resolution techniques have been proposed to increase the image resolution for face recognition. In this work, we attempt a combination of partial restoration (using super-resolution, interpolation etc.) of probe samples (long distance shots of outdoor) and simulated degradation of gallery samples (indoor shots). Due to the unavailability of any benchmark face database with gallery and probe images, we have built our own database and conducted experiments on a realistic surveillance face database. PCA and FLDA have been used as baseline face recognition classifiers. The aim is to illustrate the effectiveness of our proposed method of compensating the degradation in surveillance data, rather than designing a specific classifier space suited for degraded test probes. The efficiency of the method is shown by improvement in the face classification accuracy, while comparing results obtained separately using training with acquired indoor gallery samples and then testing with the outdoor probes.KeywordsDegradationFace databaseFace recognitionImage qualitySurveillanceSuper-resolution

Application of Face Recognition Method Under Deep Learning Algorithm in Embedded Systems