Face template creation: Is Centroid Method a suitable approach?

Deeply-learned face representations enable the success of current face recognition systems. Despite the ability of these representations to encode the identity of an individual, recent works have shown that more information is stored within, such as demographics, image characteristics, and social traits. This threatens the user's privacy, since for many applications these templates are expected to be solely used for recognition purposes. Knowing the encoded information in face templates helps to develop bias-mitigating and privacy-preserving face recognition technologies. This work aims to support the development of these two branches by analysing face templates regarding 113 attributes. Experiments were conducted on two publicly available face embeddings. For evaluating the predictability of the attributes, we trained a massive attribute classifier that is additionally able to accurately state its prediction confidence. This allows us to make more sophisticated statements about the attribute predictability. The results demonstrate that up to 74 attributes can be accurately predicted from face templates. Especially non-permanent attributes, such as age, hairstyles, haircolors, beards, and various accessories, found to be easily-predictable. Since face recognition systems aim to be robust against these variations, future research might build on this work to develop more understandable privacy preserving solutions and build robust and fair face templates.

With the development of cloud technology and the innovation of information network technology, people’s dependence on the network has gradually increased, and there are some loopholes in cloud data access. The traditional account permission model can no longer meet the conditions of cloud data access alone. If the visitor temporarily leaves the computer or goes out in an emergency, the data are likely to be leaked. Based on the importance and concern of this issue, some scholars have proposed an authentication system combined with biometric face recognition, but the traditional face recognition system has certain security risks. Such as using face pictures and videos to deceive the system, tampering with face templates, etc. Based on this, this paper proposes an encrypted face authentication system based on CNN neural network. Through the authentication of face data, the content transmission of each part is carried out in the form of ciphertext to ensure the security of information. The experimental results in this paper show that the authentication accuracy rate of DeepID is 94% when it is not encrypted, and the authentication accuracy rate decreases slightly after encryption, which is 93.3%. It is similar in other cases. When the network structure and data set remain unchanged, encryption reduces the authentication accuracy rate by 0.3%–2.4%. It can be seen that the scheme proposed in this chapter improves the system security at the cost of a smaller accuracy rate.

Face recognition is a significant area of pattern recognition and computer vision research. Illumination in face recognition is obvious yet challenging task in pattern matching. Recent researchers introduced machine learning algorithms to solve illumination problems in both indoor and outdoor scenarios. The major challenge in machine learning is the lack of classification accuracy. Thus, the novel Optimized Neural Network Algorithm (ONNA) is used to solve the aforementioned drawback. First, we propose a novel Weight Transfer Ideal Filter (WTIF) which is employed for pre-processing to remove the dark spots and shadows in an image by normalizing low frequency and high frequency of illumination. Secondly, Robust Principal Component Analysis (RPCA) is employed to perform efficient extraction of features based on image area representation. These features are given as input to ONNA which classifies the given input image under illumination. Thus we achieve the recognition of the face under various illumination conditions. Our approach is analyzed and compared with existing approaches such as Support Vector Machine (SVM) and Random Forest (RF). ONNA is better in terms of high accuracy and low error rate.

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

This paper proposes an integrated system for face recognition in cluttered images. An extension of the system is able to recognize faces by directly working with JPEG compressed image data without decompressing. The system consists of face detection using a polynomial neural network (PNN) and face recognition using pseudo-2D hidden Markov models (P2D HMM or pseudo-2D HMM). Firstly, images from a database or a Website are rescaled and preprocessed to form a standard size and file format. Secondly, face regions in image are located and extracted by PNN. Finally through face recognition the detected faces are labeled as one person of the database or rejected as new subject. Our experiments show that the face detector has a higher detection rate and low false positive rate. At the same time, the face recognition achieves higher performance with lower complexity and less computation time compared with other systems.

In this work we develop a fast facial expression recognition system based on cross correlation with low complexity by proposing a method that does not need face detection for facial points tracking. Moreover, our simple feature selection according to the facial characteristics differentiates between the six basic expressions (happiness, surprise, sadness, disgust, fear and anger). In this system, 20 selected facial feature points from the first frame to the last are tracked automatically using a cross-correlation optical flow. The extracted feature vector is then given to following classifiers: Bayes optimal classifier with two approaches in probability density function estimation, K-nearest neighbor and support vector machine with radial basis function kernel. These classifiers are analyzed according to their correct classification rate by the cross validation method. For Cohn-Kanade database the best result is obtained by Bayes optimal classifier with the average correct classification rate (Ave-CCR) of 89.67%.

Automated Face Recognition technology (AFR) has led to numerous advancements in the ever-changing world. Smart Attendance utilizes real-time Face Recognition as an actual-world solution that is integrated with everyday activities that help manage students' attendance. For this Automatic attendance system. LBPH and Haar-cascade are used to recognize and detect the image with low-rate error accuracy and for fast calculation of the image recognition. In this process the images were captured from the surveillance and recognized with low error rate.

The status of axillary lymph node metastases determines the treatment and overall survival of breast cancer (BC) patients. Three-dimensional (3D) assessment methods have advantages for spatial localization and are more responsive to morphological changes in lymph nodes than two-dimensional (2D) assessment methods, and we speculate that methods developed using 3D reconstruction systems have high diagnostic efficacy. This exploratory study included 43 patients with histologically confirmed BC diagnosed at Second Xiangya Hospital of Central South University between July 2017 and August 2020, all of whom underwent preoperative CT scans. Patients were divided into a training cohort to train the model and a validation cohort to validate the model. A 3D axillary lymph node atlas was constructed on a 3D reconstruction system to create various methods of assessing lymph node metastases for a comparison of diagnostic efficacy. Receiver operating characteristic (ROC) curve analysis was performed to assess the diagnostic values of these methods. A total of 43 patients (mean [SD] age, 47 [10] years) met the eligibility criteria and completed 3D reconstruction. An axillary lymph node atlas was established, and a correlation between lymph node sphericity and lymph node metastasis was revealed. By continuously fitting the size and characteristics of axillary lymph nodes on the 3D reconstruction system, formulas and models were established to determine the presence or absence of lymph node metastasis, and the 3D method had better sensitivity for axillary lymph node assessment than the 2D method, with a statistically significant difference in the correct classification rate. The combined diagnostic method was superior to a single diagnostic method, with a 92.3% correct classification rate for the 3D method combined with ultrasound. In addition, in patients who received neoadjuvant chemotherapy (NAC), the correct classification rate of the 3D method (72.7%) was significantly higher than that of ultrasound (45.5%) and CT (54.5%). By establishing an axillary lymph node atlas, the sphericity formula and model developed with the 3D reconstruction system achieve a high correct classification rate when combined with ultrasound or CT and can also be applied to patients receiving NAC.

Multifactor dimensionality reduction (MDR) constitutes a highly accurate classification algorithm for gene–gene interaction (GGI) identification. GGI detection quality is commonly assessed using the correct classification rate (CCR). Nevertheless, the CCR alone might not be suitable for assessing the detection of some GGIs due to various model preferences and disease complexities. Accordingly, we developed an MDR-based multiple-objective method that combines the CCR and chi-squared measures (called MDR–Cχ2) for GGI detection. In the proposed method, a Pareto set operation is executed to ensure the combination of the CCR and chi-squared measures in the MDR process for GGI detection. The most significant GGIs within the Pareto sets are determined using cross-validation consistency values. Herein, we report the MDR and MDR–Cχ2 detection success rates in a simulated environment and demonstrate that the proposed MDR–Cχ2 provides superior GGI detection success rates.

The depths, which have been used to detect outliers or to extract a representative subset, can be applied to classification. We propose a resampling-based classification method based on the fact that resampling techniques yield a consistent estimator of the distribution of a statistic. The performance of this method was evaluated with eight contaminated models in terms of Correct Classification Rates (CCRs), and the results were compared with other known methods. The proposed method consistently showed higher average CCRs and 4% higher CCR at the maximum compared to other methods. In addition, this method was applied to Berkeley data. The average CCRs were between 0.79 and 0.85.

<h3>Abstract</h3> <h3>Background</h3> With the global increase of Cesarean section delivery rates, the long-term effects of Cesarean delivery have started to become more clear. One of the most prominent complications of...

Considering the similarities between HIV-associated neurocognitive disorders (HAND) and neurodegenerative dementias and the frequency of executive dysfunctions among HIV-positive patients, we evaluated the accuracy of the Frontal Assessment Battery and Clock-Drawing Test together with the Three Questions Test and International HIV Dementia Scale to screen for HAND. A cross-sectional monocentric study was conducted from 2010 to 2017. The index tests were represented by the four screening tools; the reference standard was represented by a comprehensive neurocognitive battery used to investigate 10 cognitive domains. Patients were screened by a trained infectious diseases physician and those showing International HIV Dementia Scale scores ≤10 and/or complaining of neurocognitive symptoms were then evaluated by a trained neuropsychologist. A total of 650 patients were screened and 281 received the full neurocognitive evaluation. HAND was diagnosed in 140 individuals. The sensitivity, specificity, correct classification rate and area under the receiver operating characteristic curve (AUROC) were, respectively, as follows: Frontal Assessment Battery, 40.7%, 95.1%, 68.3% and 0.81; International HIV Dementia Scale, 74.4%, 56.8%, 65.4% and 0.73; Clock-Drawing Test, 30.9%, 73.4%, 53.8% and 0.56; and Three Questions Test, 37.3%, 54.1% and 45.7%. Raising the Frontal Assessment Battery's cut-off to ≤16 improved its sensitivity, specificity and correct classification rate to 78.0%, 63.9% and 70.8%, respectively. We observed poor screening performances of the Three Questions and Clock-Drawing Tests. While the International HIV Dementia Scale showed a poor specificity, the Frontal Assessment Battery showed the highest correct classification rate and a promising performance at different exploratory cut-offs.

In this paper, a novel method for independent component analysis (ICA) with 2-D Principle Component Analysis (2DPCA) in face recognition is presented, called 2DPCA-ICA. In this method, 2DPCA is used for dimension reduction, and ICA for recognition. As opposed to the method for ICA based on PCA in face recognition (PCA-ICA), 2DPCA is based on 2-D face image matrix, an image covariance matrix is constructed directly using 2-D image matrix, and its eigenvectors corresponding to the several larger eigenvalues are derived for whitened image matrix. It overcomes the shortcomings of PCA-ICA. To test 2DPCA-ICA and evaluate its performance, experiments are performed on Yale and ORL (Olivetti Research Laboratory) face database. Correct recognition rate of 2DPCA-ICA across all trials is higher than that of PCA-ICA and 2DPCA. Experimental results also show that features of face image extracted are more efficient by way of 2DPCA-ICA. Therefore, 2DPCA-ICA is more valid in face recognition.

At present, the correct recognition rate of face recognition algorithm is limited under unconstrained conditions. To solve this problem, a face recognition algorithm based on deep learning under unconstrained conditions is proposed in this paper. The algorithm takes LBP texture feature as the input data of deep network, and trains the network layer by layer greedily to obtain optimized parameters of network, and then uses the trained network to predict the test samples. Experimental results on the face database LFW show that the proposed algorithm has higher correct recognition rate than some traditional algorithms under unconstrained conditions. In order to further verify its effectiveness and universality, this algorithm was also tested in YALE and YALE-B, and achieved a high correct recognition rate as well, which indicated that the deep learning method using LBP texture feature as input data is effective and robust to face recognition.

548 Background: Oncotype Dx(ODX) prognosticates the risk of recurrence and predicts the benefit of adjuvant chemotherapy in estrogen-receptor-positive breast cancer (BC). However, its cost makes it prohibitive for many health care systems. Our objective is to develop a model using routine clinical and pathological parameters to estimate ODX high risk category to guide adjuvant chemotherapy decisions. Methods: We retrospectively reviewed ODX and pathology reports from 190 early BC patients (2014 to 2016) in a specialized cancer center. Variables were selected through a multiple linear regression model. Coefficients of statistically significant variables were used to build an equation. Its results were divided into 2 estimated risk categories. ODX RS was also divided into 2 categories; above or below 25 (cut-off in TAILORx and RxPONDER). The final locked model was independently validated in 57 patients. Results: Among the tested variables, tumor size (pT), progesterone receptor (PR), Ki67 and Elston-Ellis grade were significantly associated with ODX RS (Table 1). The linear predictor is: (0.2544 x pT) – (0.0739 x PR) + (0.0861 x Ki67) + (5.4232 x Elston grade). The threshold score for this equation was set on 13 (median value) to discriminate low and high estimated risks. The correct classification rate (CCR) for the training and validation sets was 60% and 56%, respectively. CCR for high risk was 72% and 87% in the training and validation sets, respectively. Conclusions: An equation based on readily available variables correctly classified more than half of cases. Although the overall CCR is modest, our equation is remarkably useful for high risk cases requiring chemotherapy. With further validation, our model could provide a clinically useful estimation of high risk at lower cost. [Table: see text]