Yale Face Dataset Research Articles

Designing face resemblance technique using near set theory under varying facial features.

Near sets (also called Descriptively Near Sets) classify nonempty sets of objects based on object feature values. The Near Set Theory provides a framework for measuring the similarity of objects based on features that describe them in much the same way humans perceive the similarity of objects. This paper presents a novel approach for face recognition using Near Set Theory that takes into account variations in facial features due to varying facial expressions, and facial plastic surgery. In the proposed work, we demonstrate two-fold usage of Near set theory; firstly, Near Set Theory as a feature selector to select the plastic surgery facial features with the help of tolerance classes, and secondly, Near Set Theory as a recognizer that uses selected prominent intrinsic facial features which are automatically extracted through the deep learning model. Extensive experimentation was performed on various facial datasets such as YALE, PSD, and ASPS. Experimentation demonstrates 93% of accuracy on the YALE face dataset, 98% of accuracy on the PSD dataset, and 98% of accuracy on the ASPS dataset. A detailed comparative analysis of the proposed work of facial resemblance with other state-of-the-art algorithms is presented in this paper. The experimentation results effectively classify face resemblance using Near Set Theory, which has outperformed several state-of-the-art classification approaches.

Fusing Supervised and Unsupervised Measures for Attribute Reduction

It is well-known that attribute reduction is a crucial action of rough set. The significant characteristic of attribute reduction is that it can reduce the dimensions of data with clear semantic explanations. Normally, the learning performance of attributes in derived reduct is much more crucial. Since related measures of rough set dominate the whole process of identifying qualified attributes and deriving reduct, those measures may have a direct impact on the performance of selected attributes in reduct. However, most previous researches about attribute reduction take measures related to either supervised perspective or unsupervised perspective, which are insufficient to identify attributes with superior learning performance, such as stability and accuracy. In order to improve the classification stability and classification accuracy of reduct, in this paper, a novel measure is proposed based on the fusion of supervised and unsupervised perspectives: (1) in terms of supervised perspective, approximation quality is helpful in quantitatively characterizing the relationship between attributes and labels; (2) in terms of unsupervised perspective, conditional entropy is helpful in quantitatively describing the internal structure of data itself. In order to prove the effectiveness of the proposed measure, 18 University of CaliforniaIrvine (UCI) datasets and 2 Yale face datasets have been employed in the comparative experiments. Finally, the experimental results show that the proposed measure does well in selecting attributes which can provide distinguished classification stabilities and classification accuracies.

Evaluating Robustness of Reconstruction Models with Adversarial Networks

With the advent of adversarial robustness as a research area, much novel work attempts to design creative defense mechanisms against adversarial vulnerabilities that arise. While classification models are the most common target of adversarial robustness research, reconstruction models are often underestimated though they play essential roles in many applications. This work evaluates reconstruction models regarding their adversarial robustness. We constructed two frameworks: a standard and a universal-attack framework. The standard framework requires an input to find its perturbation, and the universal-attack framework generates adversarial perturbation from the distribution of a dataset. Extensive experimental evidence discussed in this paper suggests that both frameworks can effectively alter how images are reconstructed and classified using classic reconstruction models trained on MNIST and Cropped Yale Face datasets. Further, these frameworks outperform state-of-the-art adversarial attacks. Moreover, we showcase using the proposed framework to retrain a reconstruction model to improve its resilience against adversarial perturbations. Furthermore, for the sake of reconstruction models, an attack may desire not to alter the latent space. Thus, we also include the analysis of the latent space.

Deep learning model for person identification and verification on low illumination images

Performances of the deep learning models for person identification and verification are degraded on low illumination images due to the generated facial embeddings being unable to be matched with the trained good illumination facial embeddings. Person verification on low illumination images is a challenging task. The existing techniques have adopted an approach of enhancing the low illumination images and performing the person verification in the enhanced images. But these techniques have not achieved satisfactory results, because the gamma value is kept constant in the power law intensity transformation function to enhance the images. To obtain better performance, we propose a deep learning-based framework that consists of a contrast enhancement module, called as contrast enhancement network (CENet); person identification; and person verification modules. The CENet is built based on the residual network, which predicts the gamma value based on the illumination of the input image. The predicted value is used to perform gamma correction on the image to improve the brightness difference between the faces and their background, whereas the existing techniques are keeping the gamma value as constant for image enhancement. After performing the image enhancement, the enhanced image is given as input to the person identification module. Then the detected faces are verified by the person verification module. Experimental results show that the proposed framework has achieved an improvement of 3.4% to 13% in person identification and verification accuracy on the extended yale face dataset to the existing methods.

Modified spider monkey optimization algorithm based feature selection and probabilistic neural network classifier in face recognition

AbstractThis paper proposes a novel and robust predictive method using modified spider monkey optimization (MSMO) and probabilistic neural network (PNN) for face recognition. The limitation of the traditional spider monkey optimization (SMO) approach to obtaining an optimal solution for classification problems is overcome by enhancing the performance of SMO by modifying the perturbation rate with a non‐linear function, thereby improving the convergence of SMO. The framework comprises image preprocessing, feature extraction using dual tree complex wavelet transform (DT‐CWT), feature selection using the modified spider monkey optimization algorithm (MSMO), and classification using PNN. The proposed method is tested on the Yale and AR Face datasets. Experimental outcomes reveal that the proposed framework attain an accuracy of 99.4% with appreciable sensitivity, specificity, and G‐mean. To examine the efficacy of MSMO, parametric studies are conducted, which showed that MSMO converges faster with high fitness when compared to similar evolutionary algorithms like Genetic Algorithm (GA), Grey Wolf Optimization Algorithm (GWO), Particle Swarm Algorithm (PSO), and Cuckoo Search (CS) in selecting the optimal feature set. The MSMO‐PNN method outperforms similar state‐of‐the‐art methods, which reveals that the method proposed is competitive. The proposed model is robust to Gaussian and salt–pepper noise, obtaining the highest accuracy of 97.89% for varied noise density and variance.

Local Triangular Coded Pattern: A Texture Descriptor for Image Classification

ABSTRACT Local binary descriptors are extensively used for image representation in many of the computer vision applications. A majority of these local binary descriptors exploit the intensity difference of the neighbouring pixels with respect to the centre pixel of the chosen region to formulate the representative value at the respective pixel position. In this paper, a novel descriptor, called Local Triangular Coded Pattern (LTCP), is introduced that utilises the relationship between a set of pixels in the triangular neighbourhood of a region to compute the descriptor. Unlike many of the other local binary descriptors, the proposed descriptor considers multiple pixels as centres within the given region to obtain the binary pattern. The performance of the LTCP descriptor is analysed by performing image classification in benchmarked texture datasets such as KTH-TIPS, Outex, Brodatz and Kylergb and in facial emotion datasets such as CK+, JAFFE, MUFE and Yale Face. The results indicate that LTCP with the Random Forest classifier gives an accuracy of 92.82%, 93.81%, 94.11% and 97.14%, respectively, on Brodatz, Outex, KTH-TIPS and Kylergb datasets for texture classification and 97.52%, 95.52%, 96.13% and 93.88%, respectively, on CK+, JAFFE, MUFE and Yale Face datasets for emotion classification. The experimental findings reflect the LTCP descriptor’s dominance and robustness over others.

Enhancing Detection Performance of Face Recognition Algorithm Using PCA-Faster R-CNN

In this paper, we present a new face detection scheme using deep learning and achieving state-of-the-art recognition performance using real-world datasets. We designed and implemented a face recognition system using Principal Component Analysis (PCA) and Faster R Convolutional Neural Network (Faster R CNN). In particular, we improve the state-of-the-art Faster RCNN framework by using Principal Component Analysis (PCA) technique and Faster R CNN to detect and recognise faces in a face database. The Principal Component Analysis (PCA) was used to extract features and dimensionality reduction from the face database, while the Faster R Convolutional Neural Network algorithm was used to identify patterns in the dataset via training the neural network. The three real-world datasets used in our experiment are ORL, Yale, and California face dataset. When implemented on the ORL face dataset, the algorithm achieved average recognition accuracy of 99%, with a recognition time of 147.72 seconds for 10 runs, and the recognition time/image was 0.3 sec/image on 400 images. The Yale face dataset achieved average recognition accuracy of 99.24% with a recognition time of 63.45 seconds for 10 runs, and the recognition time/image was 0.53 sec/image on 120 images. Finally, on California Face Database (CFD), it achieved average recognition accuracy of 99.52% with a recognition time of 226.05 seconds for 10 runs, and the recognition time/image was 0.27 sec/image on 827 images. On the CFD dataset, however, the proposed approach has excellent classification performance when the recall ratio is high. The proposed method achieves a higher recall and accuracy ratio than the Faster RCNN without PCA method. For the F-score, the proposed method achieved 0.98, which is significantly higher than the 0.95 achieved by the Faster-RCNN. This demonstrates the superiority of our model performance-wise as against state-of-the-art, both in terms of accuracy and fast recognition. Therefore our model is more efficient when compared to the latest researches done in the area of facial recognition.

Experimental Quantum Principal Component Analysis via Parametrized Quantum Circuits.

Principal component analysis (PCA) is a widely applied but rather time-consuming tool in machine learning techniques. In 2014, Lloyd, Mohseni, and Rebentrost proposed a quantum PCA (qPCA) algorithm [Lloyd, Mohseni, and Rebentrost, Nat. Phys. 10, 631 (2014)NPAHAX1745-247310.1038/nphys3029] that still lacks experimental demonstration due to the experimental challenges in preparing multiple quantum state copies and implementing quantum phase estimations. Here, we propose a new qPCA algorithm using the hybrid classical-quantum control, where parameterized quantum circuits are optimized with simple measurement observables, which significantly reduces the experimental complexity. As one important PCA application, we implement a human face recognition process using the images from the Yale Face Dataset. By training our quantum processor, the eigenface information in the training dataset is encoded into the parameterized quantum circuit, and the quantum processor learns to recognize new face images from the test dataset with high fidelities. Our work paves a new avenue toward the study of qPCA applications in theory and experiment.

Bayesian Low Rank Tensor Ring for Image Recovery.

Low rank tensor ring based data recovery can recover missing image entries in signal acquisition and transformation. The recently proposed tensor ring (TR) based completion algorithms generally solve the low rank optimization problem by alternating least squares method with predefined ranks, which may easily lead to overfitting when the unknown ranks are set too large and only a few measurements are available. In this article, we present a Bayesian low rank tensor ring completion method for image recovery by automatically learning the low-rank structure of data. A multiplicative interaction model is developed for low rank tensor ring approximation, where sparsity-inducing hierarchical prior is placed over horizontal and frontal slices of core factors. Compared with most of the existing methods, the proposed one is free of parameter-tuning, and the TR ranks can be obtained by Bayesian inference. Numerical experiments, including synthetic data, real-world color images and YaleFace dataset, show that the proposed method outperforms state-of-the-art ones, especially in terms of recovery accuracy.

Face Recognition using Color and Edge Orientation Difference Histogram

Face recognition is a challenging problem because of different illuminations, poses, facial expressions, and occlusions. In this paper, a new robust face recognition method is proposed based on color and edge orientation difference histogram. Firstly, color and edge orientation difference histogram is extracted using color, color difference, edge orientation and edge orientation difference of the face image. Then, backward feature selection is employed to reduce the number of features. Finally, Canberra measure is used to assess the similarity between the images. Color and edge orientation difference histogram shows uniform color difference and edge orientation difference between two neighboring pixels. This histogram will be effective for face recognition due to different skin colors and different edge orientations of the face image, which leads to different light reflection. The proposed method is evaluated on Yale and ORL face datasets. These datasets are consisted of gray-scale face images under different illuminations, poses, facial expressions and occlusions. The recognition rate over Yale and ORL datasets is achieved 100% and 98.75% respectively. Experimental results demonstrate that the proposed method outperforms the existing methods in face recognition.

Principal component analysis, hidden Markov model, and artificial neural network inspired techniques to recognize faces

AbstractFace Recognition is a challenging task for recognizing and detecting the identity of an individual. Although, plethora of work has already been done in the field of pattern recognition still there has been lot which has not been addressed in any of the literature. In the current research, we have presented a comparative analysis using three popularly known techniques for face recognition namely, Principal Components Analysis (PCA) using Eigen Faces, Hidden Markov Model (HMM) using Singular Value Decomposition, and Artificial Neural Network (ANN) using Gabor filters. These techniques are implemented and evaluated using various measuring metrics such as false acceptance, false recognition rate, and so on. We used ORL and Yale Face dataset to test the robustness of implemented algorithms. Results show that ANN model for face recognition outperforms the other two techniques by achieving more accurate results and shows the highest recognition rate of 97.49% on ORL database. Moreover, it is also observed that ANN model shows the minimum error count of about 2.502% on ORL database while it is 3.5% on Yale Face dataset. To evaluate further, the implemented techniques are compared with best known techniques in class implemented by various researchers.

Genetic‐based feature fusion in face recognition using arithmetic coded local binary patterns

Local binary patterns (LBPs) are one of the attempts for gathering local features with face recognition algorithms. Although the application of LBP's in many recognition contents is too apparent, these methods have limited accuracy because of their threshold value. One problem is earning one value for two different regions with a diverse pixel neighbourhood, which causes mistakes in feature vector and decreases the discriminative power. In this study, the authors proposed a modified LBP that covers the LBP's disadvantages. The proposed approach is arithmetic coded LBP (ACLBP) that uses arithmetic coding process during LBP calculation instead of applying original thresholds. The proposed policy addresses the problem of returning one similar LBP value for two different patches. Moreover, the proposed method modifies LBP by using a different threshold for calculating the pixels differences. Using this algorithm, the authors conducted a genetic-based feature fusion method by combining LBP and histogram of oriented gradients and ACLBP. The proposed approach could work better on LFW dataset, and the ORL dataset and Yale face dataset that shows the improving role of ACLBP in comparison with the earlier version of LBP.

Local improvement approach and linear discriminant analysis-based local binary pattern for face recognition

Face recognition applications focus on local features to prevent detailed information from being omitted while the feature extraction processes. This paper is based on presenting a local pattern-based model to extract more discriminative features that lead to more accurate classification. In local pattern-based feature extraction, the LBP is one of the most important approaches that many variants of this method have been proposed till now. LBP calculation is based on differences between the central pixel and the desired one. In contrast, the information hidden in the selected pixel’s neighborhood pixels is not included in this process. This paper proposes the DR_LBP approach to address this failure by defining distances and using some of them in a ratio form. Successful results have been earned in many experimental results. In LBP, the calculations’ primary flow takes advantage of two pixels in the LBP box, the central and the desired pixel. Contrary to the original LBP, this paper’s proposed approach uses three pixels of LBP box to conduct the feature vector, which leads to employing the information hidden in the relationship between neighboring pixels. This approach applies the experiments on two standard datasets, ORL Yale face and Faces94 dataset. The accuracy percent of the proposed plan is 95.95, 94.09 and 98.01 on ORL, Yale face and Faces94 dataset, respectively, which is the reason to present this model as a new face feature extraction approach.

Robust sparse principal component analysis by DC programming algorithm

The classical principal component analysis (PCA) is not sparse enough since it is based on the L2-norm that is also prone to be adversely affected by the presence of outliers and noises. In order to address the problem, a sparse robust PCA framework is proposed based on the min of zero-norm regularization and the max of Lp-norm (0 < p ≤ 2) PCA. Furthermore, we developed a continuous optimization method, DC (difference of convex functions) programming algorithm (DCA), to solve the proposed problem. The resulting algorithm (called DC-LpZSPCA) is convergent linearly. In addition, when choosing different p values, the model can keep robust and is applicable to different data types. Numerical simulations are simulated in artificial data sets and Yale face data sets. Experiment results show that the proposed method can maintain good sparsity and anti-outlier ability.

Two-dimensional joint local and nonlocal discriminant analysis-based 2D image feature extraction for deep learning

This paper proposes a new two-dimensional manifold learning algorithm called two-dimensional joint local/nonlocal discriminant analysis (2DJLNDA) for 2D image feature extraction, which directly extracts projective vectors from 2D image matrices rather than image vectors. Different from other typical 2D methods, e.g., two-dimensional principal component analysis (2DPCA), two-dimensional linear discriminative analysis (2DLDA), two-dimensional locality-preserving projection (2DLPP), 2DJLNDA preserves not only local/nonlocal intrinsic structure but also local/nonlocal penalization structure of the image data in the high-dimensional space, which can be powerful in extracting intrinsic information of the image data in the low-dimensional space. The experimental results on the ORL, Yale, AR and UMIST face datasets indicate that 2DJLNDA is capable of extracting effective image features and outperforms 2DPCA, 2DLDA and 2DLPP. The 2D image features extracted by 2DJLNDA further improve the performance of deep neural networks (DNNs), e.g., stacked denoising autoencoder, and convolutional neural network (CNN) significantly. These studying results illustrate that the feature face images will provide more discriminant features than the original face images for DNNs. Therefore, 2DJLNDA-based 2D feature image extraction can be used as an effective preprocessing of DNNs (e.g., CNN) for face recognition.

Cross-covariance regularized autoencoders for nonredundant sparse feature representation

Abstract We propose a new feature representation algorithm using cross-covariance in the context of deep learning. Existing feature representation algorithms based on the sparse autoencoder and nonnegativity-constrained autoencoder tend to produce duplicative encoding and decoding receptive fields, which leads to feature redundancy and overfitting. We propose using the cross-covariance to regularize the feature weight vector to construct a new objective function to eliminate feature redundancy and reduce overfitting. The results from the MNIST handwritten digits dataset, the NORB normalized-uniform dataset and the Yale face dataset indicate that relative to other algorithms based on the conventional sparse autoencoder and nonnegativity-constrained autoencoder, our method can effectively eliminate feature redundancy, extract more distinctive features, and improve sparsity and reconstruction quality. Furthermore, this method improves the image classification performance and reduces the overfitting of conventional networks without adding more computational time.

Local receptive fields based extreme learning machine with hybrid filter kernels for image classification

In this paper, an innovative method called extreme learning machine with hybrid local receptive fields (ELM-HLRF) is presented for image classification. In this method, filters generated by Gabor functions and the randomly generated convolution filters are incorporated into the convolution filter kernels of local receptive fields based extreme learning machine (ELM-LRF). Extreme learning machine (ELM) is derived from single hidden layer feed-forward neural networks, and the parameters of its hidden layer can be generated randomly. As locally connected ELM, ELM-LRF directly processes information with strong correlations such as images and speech. In this paper, two main contributions are proposed to improve the classification performance of ELM-LRF. First, the Gabor functions are used as one kind of convolution filter kernels of ELM-HLRF to execute image classification. Second, we use a data augmentation method to preprocess training images to avoid overfitting. Experiments on the Outex texture dataset, the Yale face dataset, the ORL face database and the NORB dataset demonstrate that ELM-HLRF outperforms ELM-LRF, ELM and support vector machine in classification accuracy, and the presented data augmentation method improves the classification performance.

Robust $$l_{2,1}$$ l 2 , 1 Norm-Based Sparse Dictionary Coding Regularization of Homogenous and Heterogenous Graph Embeddings for Image Classifications

In the field of manifold learning, Marginal Fisher Analysis (MFA), Discriminant Neighborhood Embedding (DNE) and Double Adjacency Graph-based DNE (DAG-DNE) construct the graph embedding for homogeneous and heterogeneous k-nearest neighbors (i.e. double adjacency) before feature extraction. All of them have two shortcomings: (1) vulnerable to noise; (2) the number of feature dimensions is fixed and likely very large. Taking advantage of the sparsity effect and de-noising property of sparse dictionary, we add the $$l_{2,1}$$ norm-based sparse dictionary coding regularization term to the graph embedding of double adjacency, to form an objective function, which seeks a small amount of significant dictionary atoms for feature extraction. Since our initial objective function cannot generate the closed-form solution, we construct an auxiliary function instead. Theoretically, the auxiliary function has closed-form solution w.r.t. dictionary atoms and sparse coding coefficients in each iterative step and its monotonously decreased value can pull down the initial objective function value. Extensive experiments on the synthetic dataset, the Yale face dataset, the UMIST face dataset and the terrain cover dataset demonstrate that our proposed algorithm has the ability of pushing the separability among heterogenous classes onto much fewer dimensions, and robust to noise.

Nonredundant sparse feature extraction using autoencoders with receptive fields clustering

This paper proposes new techniques for data representation in the context of deep learning using agglomerative clustering. Existing autoencoder-based data representation techniques tend to produce a number of encoding and decoding receptive fields of layered autoencoders that are duplicative, thereby leading to extraction of similar features, thus resulting in filtering redundancy. We propose a way to address this problem and show that such redundancy can be eliminated. This yields smaller networks and produces unique receptive fields that extract distinct features. It is also shown that autoencoders with nonnegativity constraints on weights are capable of extracting fewer redundant features than conventional sparse autoencoders. The concept is illustrated using conventional sparse autoencoder and nonnegativity-constrained autoencoders with MNIST digits recognition, NORB normalized-uniform object data and Yale face dataset.

Adaptive stochastic gradient descent on the Grassmannian for robust low‐rank subspace recovery

In this study, the authors present GASG21 (Grassmannian adaptive stochastic gradient for L 2,1 norm minimisation), an adaptive stochastic gradient algorithm to robustly recover the low-rank subspace from a large matrix. In the presence of column outliers corruption, the authors reformulate the classical matrix L 2,1 norm minimisation problem as its stochastic programming counterpart. For each observed data vector, the low-rank subspace S is updated by taking a gradient step along the geodesic of Grassmannian. In order to accelerate the convergence rate of the stochastic gradient method, the authors choose to adaptively tune the constant step-size by leveraging the consecutive gradients. Numerical experiments on synthetic data and the extended Yale face dataset demonstrate the efficiency and accuracy of the proposed GASG21 algorithm even with heavy column outliers corruption.