Search

No-reference (NR) image quality assessment (IQA) is an important task of computer vision. Most NR-IQA methods via deep neural networks do not reach desirable IQA performance and have bulky models which make them difficult to be used in the practical scenarios. This paper proposes a lightweight transformer and multi-head prediction network for NR-IQA. The proposed method consists of two lightweight modules: feature extraction and multi-head prediction. The module of feature extraction exploits lightweight transformer blocks to learn features at different scales for measuring different image distortions. The module of multi-head prediction uses three weighted prediction blocks and an FC layer to aggregate the learned features for predicting image quality score. The weighted prediction block can measure the importance of different elements of input feature at the same scale. Since the importance of feature elements at the same scale and the importance of the features at different scales are both considered, the module of multi-head prediction can provide more accurate prediction results. Extensive experiments on the standard IQA datasets are conducted. The results show that the proposed method outperforms some baseline NR-IQA methods in IQA performance on the large image datasets. For the model complexity, the proposed method is also superior to several recent NR-IQA methods. [ABSTRACT FROM AUTHOR]

The convolutional neural network has been proven to be a robust recognition and diagnosis model for modelling diseases. In general, a convolutional neural network consists of feature learning which is considered a vital process in terms of feature extraction and classifier algorithms used for decision-making. In recent years, various pooling operations considered the spine of the feature learning layer have been developed in the literature; unfortunately, few of them can be spread to tackle the overfitting problem. In this paper, a novel pooling approach termed whole image average pooling is proposed to hold over the most significant feature information on the learning layer. The proposed method consists of two main stages: in the first stage, the average of all the values of the input image is calculated, in the second stage, the input image is divided into regions that represent the feature set of the input image, and then, the values of the regions are calculated. While obtaining the values of these features, the average of all values of the input image is taken into account. The proposed whole image average pooling approach addresses the overfitting problem by processing all pixel values in the image data and is used to get the important feature information from the medical datasets. To evaluate the performance of the proposed whole image average pooling, four clinically applicable tumour datasets as medical datasets have been trained and tested on seven convolutional neural network models AlexNet, VGG16, ResNet18, GoogleNet, ResNet50, R-CNN, and Faster R-CNN. The experimental outcomes demonstrate that the proposed whole image average pooling consistently outperformed the state-of-the-art pooling operation in terms of accuracy, sensitivity, specificity, positive predictive value, false-positive rate, and F1-measure evaluation criteria in three datasets. [ABSTRACT FROM AUTHOR]

In the previous works on heat stress detection, various state-of-the-art machine learning techniques had been used to detect stress patterns from electroencephalographic (EEG) signals. Since the handcrafted feature engineering-based approaches pose certain limitations and are sensitive to transform the nonlinearities of EEG, deep learning techniques have drawn attention in the domain of EEG-based applications. Moreover, existing approaches for stress detection consider the whole frequency band (delta to gamma) which conceals the redundant and lossy information that increased the false detection rate. Also, these approaches were implemented only to identify the stress in binary classes and confined their application to identifying the level of stress. Therefore, a multi-scale channel-wise convolution-based multi-level heat stress assessment has been proposed in this paper. The novel contributions of this work are (1) designing a multi-scale convolutional neural network (MS-CNN) for extracting precise information from individual frequency bands of EEG, (2) use of sparse connectivity to reduce the redundant information and increase the performance with less trainable parameters, and (3) multi-level heat stress assessment for precise identification of the severity of stress. The proposed approaches were evaluated on pre-recorded data of 10 rats in a simulated laboratory environment. The high accuracy of approximately 96–97% and 90–92% has been achieved for binary and multi-level stress detection. There is approximately a 6 to 50% reduction in the trainable parameters with a 2% improvement of accuracy achieved on adopting channel-wise convolution over MS-CNN and deep convolutional neural network (DCNN). This demonstrates the effectiveness of the adopted multiscale feature extraction and sparse connectivity to improve the performance with a less complex model. [ABSTRACT FROM AUTHOR]

Diabetic Retinopathy (DR), one of the most progressive sight-threatening diseases caused by the long-term diabetic condition, can lead to vision impairment and blindness later. Early diagnosis and timely treatment help control and avert DR from its progression. However, manual grading is exceptionally challenging and arduous due to the complex anatomical features in the retina. Therefore, developing an automated diagnostic method for screening DR is obligatory. This paper proposes a deep learning-enabled optimized feature selection approach to classify the stage of DR severity in the fundus image. At first, a pre-processing phase eradicates the noise and improves the contrast in the retinal fundus image. Then, blood vessel segmentation is performed using the Coherence Enhancing Energy Based Regularized Level Set Evolution method in the green channel fundus image. Subsequently, the optic disk is segmented with Canny Anisotropic Diffusion filter and morphological transformations. Next, the candidate lesion region is detected using an Attention-based Fusion Network (AFU-Net). Then, shape and texture features are extracted, and then, the optimal subset of features is selected using the Improved Harris Hawk Optimization algorithm. Finally, a deep Convolutional Neural Network classifies the DR stages, and the model weight is updated using the same algorithm. The proposed method achieved superior performance in two benchmark public datasets compared with the existing state-of-the-art methods using F1-score, accuracy, sensitivity, and specificity measures. [ABSTRACT FROM AUTHOR]

In recent years, the Internet of Things (IoT) has received a lot of attention. It has been used in many applications such as the control industry, industrial plants, and medicine. In this regard, a fundamental necessity is to implement security in IoT. To this end, Network intrusion detection systems (NIDSs) have been recently in the detection of network attacks and threats. Currently, these systems use a variety of deep learning (DL) models such as the convolutional neural networks to improve the detection of attacks. However, almost all current DL-based NIDSs are made up of many layers, and therefore, they need a lot of processing resources because of their high number of parameters. On the other hand, due to the lack of processing resources, such inefficient DL models are unusable in IoT devices. This paper presents a very accurate NIDS that is named DFE, and it uses a very lightweight and efficient neural network based on the idea of deep feature extraction. In this model, the input vector of the network is permuted in a 3D space, and its individual values are brought close together. This allows the model to extract highly discriminative features using a small number of layers without the need to use large 2D or 3D convolution filters. As a result, the network can achieve an accurate classification using a significantly small number of required calculations. This makes the DFE ideal for real-time intrusion detection by IoT devices with limited processing capabilities. The efficacy of the DFE has been evaluated using three popular public datasets named UNSW-NB15, CICIDS2017, and KDDCup99, and the results show the superiority of the proposed model over the state-of-the-art algorithms [ABSTRACT FROM AUTHOR]

In this paper, we investigate Hebbian learning strategies applied to Convolutional Neural Network (CNN) training. We consider two unsupervised learning approaches, Hebbian Winner-Takes-All (HWTA), and Hebbian Principal Component Analysis (HPCA). The Hebbian learning rules are used to train the layers of a CNN in order to extract features that are then used for classification, without requiring backpropagation (backprop). Experimental comparisons are made with state-of-the-art unsupervised (but backprop-based) Variational Auto-Encoder (VAE) training. For completeness,we consider two supervised Hebbian learning variants (Supervised Hebbian Classifiers—SHC, and Contrastive Hebbian Learning—CHL), for training the final classification layer, which are compared to Stochastic Gradient Descent training. We also investigate hybrid learning methodologies, where some network layers are trained following the Hebbian approach, and others are trained by backprop. We tested our approaches on MNIST, CIFAR10, and CIFAR100 datasets. Our results suggest that Hebbian learning is generally suitable for training early feature extraction layers, or to retrain higher network layers in fewer training epochs than backprop. Moreover, our experiments show that Hebbian learning outperforms VAE training, with HPCA performing generally better than HWTA. [ABSTRACT FROM AUTHOR]

Although convolutional neural networks have achieved success in the field of image classification, there are still challenges in the field of agricultural product quality sorting such as machine vision-based jujube defects detection. The performance of jujube defect detection mainly depends on the feature extraction and the classifier used. Due to the diversity of the jujube materials and the variability of the testing environment, the traditional method of manually extracting the features often fails to meet the requirements of practical application. In this paper, a jujube sorting model in small data sets based on convolutional neural network and transfer learning is proposed to meet the actual demand of jujube defects detection. Firstly, the original images collected from the actual jujube sorting production line were pre-processed, and the data were augmented to establish a data set of five categories of jujube defects. The original CNN model is then improved by embedding the SE module and using the triplet loss function and the center loss function to replace the softmax loss function. Finally, the depth pre-training model on the ImageNet image data set was used to conduct training on the jujube defects data set, so that the parameters of the pre-training model could fit the parameter distribution of the jujube defects image, and the parameter distribution was transferred to the jujube defects data set to complete the transfer of the model and realize the detection and classification of the jujube defects. The classification results are visualized by heatmap through the analysis of classification accuracy and confusion matrix compared with the comparison models. The experimental results show that the SE-ResNet50-CL model optimizes the fine-grained classification problem of jujube defect recognition, and the test accuracy reaches 94.15%. The model has good stability and high recognition accuracy in complex environments. [ABSTRACT FROM AUTHOR]

Recent advancements in multielectrode methods and spike-sorting algorithms enable the in vivo recording of the activities of many neurons at a high temporal resolution. These datasets offer new opportunities in the investigation of the biological neural code, including the direct testing of specific coding hypotheses, but they also reveal the limitations of present decoder algorithms. Classical methods rely on a manual feature extraction step, resulting in a feature vector, like the firing rates of an ensemble of neurons. In this paper, we present a recurrent neural-network-based decoder and evaluate its performance on experimental and artificial datasets. The experimental datasets were obtained by recording the auditory cortical responses of rats exposed to sound stimuli, while the artificial datasets represent preset encoding schemes. The task of the decoder was to classify the action potential timeseries according to the corresponding sound stimuli. It is illustrated that, depending on the coding scheme, the performance of the recurrent-network-based decoder can exceed the performance of the classical methods. We also show how randomized copies of the training datasets can be used to reveal the role of candidate spike-train features. We conclude that artificial neural network decoders can be a useful alternative to classical population vector-based techniques in studies of the biological neural code. [ABSTRACT FROM AUTHOR]

Ball mills are extensively used in the size reduction process of different ores and minerals. The fill level inside a ball mill is a crucial parameter which needs to be monitored regularly for optimal operation of the ball mill. In this paper, a vibration monitoring-based method is proposed and tested for estimating the fill level inside a laboratory-scale ball mill. A vibration signal is captured from the base of a laboratory-scale ball mill by using a ± 5 g accelerometer. Features are extracted from the vibration signal by using different transforms such as fast Fourier transform, discrete wavelet transform, wavelet packet decomposition, and empirical mode decomposition. These features are given as input to an artificial neural network which is used to predict the percentage fill level inside the ball mill. In this paper, the predicted fill level obtained by using different features are compared. It is found that the predicted fill level due to features obtained after fast Fourier transform outperforms other transforms. [ABSTRACT FROM AUTHOR]

In this study, efficient deep transfer learning models are proposed to classify six types of volcanic rocks, and this paper has a novelty in classifying volcanic rock types for the first time using thin section images. Convolutional neural network-based DenseNet121 and ResNet50 networks, which are transfer learning methods, are used to extract the features from thin section images of rocks, and the classification process is carried out with a single-layer fully connected neural network. The proposed models are trained and tested on 1200 thin section images using four different optimizers (Adadelta, ADAM, RMSprop, SGD). AUC, accuracy, precision, recall and f1-score are used as performance metrics. Proposed models are run 10 times for each optimizer. DenseNet121 classifies volcanic rock types using RMSprop with an average accuracy of 99.50% and a maximum of 100.00%, and ResNet50 classifies using ADAM with an average accuracy of 98.80% and a maximum of 99.72%. Thus, the applied deep transfer learning is promising in geosciences and can be used to identify rock types quickly and accurately. [ABSTRACT FROM AUTHOR]

Wearable technology offers a prospective solution to the increasing demand for activity monitoring in pervasive healthcare. Feature extraction and selection are crucial steps in activity recognition since it determines the accuracy of activity classification. However, existing feature extraction and selection methods involve manual feature engineering, which is time-consuming, laborious and prone to error. Therefore, this paper proposes an unsupervised feature learning method that automatically extracts and selects the features without human intervention. Specifically, the proposed method jointly trains a convolutional denoising autoencoder with a convolutional neural network to learn the underlying features and produces a compact feature representation of the data. This allows not only more accurate and discriminative features to be extracted but also reduces the computational cost and improves generalization of the classification models. The proposed method was evaluated and compared with deep learning convolutional neural networks on a public dataset. Results have shown that the proposed method can learn a salient feature representation and subsequently recognize the activities with an accuracy of 0.934 and perform comparably well to the convolutional neural networks. [ABSTRACT FROM AUTHOR]

Hand pose tracking is essential in sign languages. An automatic recognition of performed hand signs facilitates a number of applications, especially for people with speech impairment to communication with normal people. This framework which is called ASLNN proposes a new hand posture recognition technique for the American sign language alphabet based on the neural network which works on the geometrical feature extraction of hands. A user's hand is captured by a three-dimensional depth-based sensor camera; consequently, the hand is segmented according to the depth analysis features. The proposed system is called depth-based geometrical sign language recognition as named DGSLR. The DGSLR adopted in easier hand segmentation approach, which is further used in segmentation applications. The proposed geometrical feature extraction framework improves the accuracy of recognition due to unchangeable features against hand orientation compared to discrete cosine transform and moment invariant. The findings of the iterations demonstrate the combination of the extracted features resulted to improved accuracy rates. Then, an artificial neural network is used to drive desired outcomes. ASLNN is proficient to hand posture recognition and provides accuracy up to 96.78% which will be discussed on the additional paper of this authors in this journal. [ABSTRACT FROM AUTHOR]

Brain–computer interface (BCI) is the current trend in technology expansion as it provides an easy interface between human brain and machine. The demand for BCI-based applications is growing tremendously, and efforts are in progress to deploy BCI devices for real-world applications. One of the widely known applications of BCI technology is rehabilitation in which BCI devices can provide various types of assistance to specially abled persons. In this paper, the effect of hand actions on objects is analyzed for motor-related mental task. The proposed approach analyzes electroencephalogram (EEG)-based brain activity which was captured for images shown with different gripping actions on objects. The EEG recordings are first pre-processed, followed by extraction of epochs and frequency bands using discrete wavelet transform; afterward, feature extraction followed by training and classification steps is performed for classifying the grip action into congruent (correct) and incongruent (incorrect) grip categories. The proposed work makes use of average power and relative wavelet energy as discriminating features which are then fed to train an artificial neural network for automatically classifying the incoming EEG patterns into correct or incorrect object hand grips. The performance evaluation of the proposed system is done on real EEG dataset obtained from 14 subjects. Experimental results have shown an accuracy of 75%. Also, to evaluate the effectiveness of our work, a comparison of our work with other state-of-the-art works reported by different authors is presented at the end. The results show the effectiveness of proposed approach and suggest further that the system can be used to analyze and train subjects having motor-related disabilities for perceiving correct or incorrect hand grips on objects. [ABSTRACT FROM AUTHOR]

Visual odometry (VO) is a fundamental and challenging problem in both the computer vision community and the robotics community. VO refers the process of recovering the relative movements of a camera by analyzing the associated image sequence. While VO is generally formulated as descriptors-based feature tracking with outliers rejection and global optimization, these algorithms are not only computationally expensive but also lack robustness. In the paper, a biologically inspired solution to the monocular visual odometry problem was presented, which was named as shunting short-term memory monocular visual odometry. The proposed method is simple and concise in both concept and implementation. To be more specific, it utilizes the shunting short-term memory to represent the key frames and the latest observations and also to adapt to uncertainties and ambiguities. And scan matching scheme is adopted to search the movement that best explained the difference between the latest observation and the key frame. Because of the dynamic properties of the neural network, the proposed method requires neither explicit extraction of features and descriptors, nor outliers detection and bundle optimization. Theoretical analysis in the paper showed that the proposed method has Lyapunov stability and constant computational complexity. The proposed method was also compared with the classical monocular VO algorithm in real indoor environments, and the experimental results proved that the proposed method outperforms the classical method on both effectiveness and robustness. [ABSTRACT FROM AUTHOR]

In many classification problems, it is necessary to consider the specific location of an n-dimensional space from which features have been calculated. For example, considering the location of features extracted from specific areas of a two-dimensional space, as an image, could improve the understanding of a scene for a video surveillance system. In the same way, the same features extracted from different locations could mean different actions for a 3D HCI system. In this paper, we present a self-organizing feature map able to preserve the topology of locations of an n-dimensional space in which the vector of features have been extracted. The main contribution is to implicitly preserving the topology of the original space because considering the locations of the extracted features and their topology could ease the solution to certain problems. Specifically, the paper proposes the n-dimensional constrained self-organizing map preserving the input topology (nD-SOM-PINT). Features in adjacent areas of the n-dimensional space, used to extract the feature vectors, are explicitly in adjacent areas of the nD-SOM-PINT constraining the neural network structure and learning. As a study case, the neural network has been instantiate to represent and classify features as trajectories extracted from a sequence of images into a high level of semantic understanding. Experiments have been thoroughly carried out using the CAVIAR datasets (Corridor, Frontal and Inria) taken into account the global behaviour of an individual in order to validate the ability to preserve the topology of the two-dimensional space to obtain high-performance classification for trajectory classification in contrast of non-considering the location of features. Moreover, a brief example has been included to focus on validate the nD-SOM-PINT proposal in other domain than the individual trajectory. Results confirm the high accuracy of the nD-SOM-PINT outperforming previous methods aimed to classify the same datasets. [ABSTRACT FROM AUTHOR]

Artificial metaplasticity is the machine learning algorithm inspired in the biological metaplasticity of neural synapses. Metaplasticity stands for plasticity of plasticity, and as long as plasticity is related to memory, metaplasticity is related to learning. Implemented in supervised learning assuming input patterns distribution or a related function, it has proved to be very efficient in performance and in training convergence for multidisciplinary applications. Now, for the first time, this kind of artificial metaplasticity is implemented in an unsupervised neural network, achieving also excellent results that are presented in this paper. To compare results, a modified self-organization map is applied to the classification of MIT-BIH cardiac arrhythmias database. [ABSTRACT FROM AUTHOR]

Fingerprint pattern recognition and classification can be of assistance in the research on human personality. In some previous studies, fingerprints were classified into four categories to speed up recognition, but the method of that classification is not suitable for researching the diversity of human personalities. Therefore, in this paper, fingerprint patterns were classified into six types and the accuracy of the recognition was improved to facilitate the research on human personality characteristics. Based on this idea, a six-category fingerprint database is annotated manually and a convolutional neural network (CNN) is proposed for identifying real fingerprint patterns. The new CNN consists of four convolutional layers, three max-pooling layers, two norm layers, and three fully connected layers. The best accuracy the model achieved was 94.87% for a six-category fingerprint database and 92.9% accuracy for a four-category fingerprint database. The results of experimental tests show that the proposed model can recognize the pattern features from a large fingerprint database using the automatic learning and feature extraction abilities of the CNN to get a greater accuracy than in previous experiments. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

In the field of artificial intelligence, the development of many technologies requires technical support for relational classification. Recently, deep learning has been applied more and more to text-based entity relationship classification tasks, but most of the previous methods need to use syntax or dependency structure feature. However, due to the time and space complexity of syntactic parsing, the structural features are inconvenient to use directly in the pre-processing stage. In addition, structural features may have serious domain dependence problems. This paper studies the current recommendation algorithm, analyzes the current research status of the recommendation system, and deeply analyzes the research of deep learning in the field of recommendation systems, based on BPSO algorithm, the context complex segmentation method is applied, and then the deep convolutional neural network is applied for feature extraction. The extracted feature set is sent to WordEmbedding, and using the technology to generates the word vector, the input layer of the CBOW is used to represent the size of the training window. The experimental results show that the model has obvious advantages over the methods proposed in other literature. It can adapt to multi-category context semantic analysis, more accurate related recommendations, and obtain a better user experience. [ABSTRACT FROM AUTHOR]

Human physical action classification is an emerging area of research for human-to-machine interaction, which can help to disable people to interact with real world, and robotics application. EMG signals measure the electrical activity muscular systems, which involved in physical action of human. EMG signals provide more information related to physical action. In this paper, we proposed deep transfer learning-based approach of human action classification using surface EMG signals. The surface EMG signals are represented by time–frequency image (TFI) by using short-time Fourier transform. TFI is used as input to pre-trained convolutional neural network models, namely AlexNet and VGG16, for deep feature extraction, and support vector machine (SVM) classifier is used for classification of physical action of EMG signals. Also, the fine-tuning of the pre-trained AlexNet model is also considered. The experimental results show that deep feature extraction and SVM classification method and fine-tuning have obviously improved the classification accuracy when compared with various results from the literature. The 99.04% accuracy score is obtained with AlexNet fc6 + AlexNet fc7 + VGG16 fc6 + VGG16 fc7 deep feature concatenation and SVM classification. 98.65% accuracy score is performed by fine-tuning of the AlexNet model. We also compare the obtained results with some of the existing methods. The comparisons show that the deep feature concatenation and SVM classification method provide better classification accuracy than the compared methods. [ABSTRACT FROM AUTHOR]

As the most direct media form in the Internet age, text information has received more and more attention in the era of the maturity of social networks and the rapid development of mobile terminals technology (general name). In this paper, the methods of removing invalid records, manually filling in vacancy values and using global constant to fill in vacancy values are used to preprocess semantic ontology. Then, the semantic ontology is further processed by using Chinese word segmentation to reduce the computational cost of later feature extraction and to increase the effectiveness of feature extraction. Using X2 statistic CHI(t , x) to extract the feature of ontology semantics, the high-dimensional data are mapped to low-dimensional space by transformation, which reduces the dimension of vector space and simplifies the calculation. Finally, using news data source and MATLAB software design platform to do classification experiment using convolutional neural network classification and from recall rate, correct rate and F-measure, three indicators to compare the classification effect of NB, KNN, SVM, the effectiveness of the experimental method is verified. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel writer-independent global feature extraction framework for the task of automatic signature verification which aims to make robust systems for automatically distinguishing negative and positive samples. Our method consists of an autoencoder for modeling the sample space into a fixed-length latent space and a siamese network for classifying the fixed-length samples obtained from the autoencoder based on the reference samples of a subject as being genuine or forged. During our experiments, usage of attention mechanism and applying downsampling significantly improved the accuracy of the proposed framework. We evaluated our proposed framework using SigWiComp2013 Japanese and GPDSsyntheticOnLineOffLineSignature datasets. On the SigWiComp2013 Japanese dataset, we achieved 8.65% equal error rate (EER) that means 1.2% relative improvement compared to the best-reported result. Furthermore, on the GPDSsyntheticOnLineOffLineSignature dataset, we achieved average EERs of 0.13%, 0.12%, 0.21% and 0.25%, respectively, for 150, 300, 1000 and 2000 test subjects which indicate improvement in relative EER on the best-reported result by 95.67%, 95.26%, 92.9% and 91.52%, respectively. Apart from the accuracy gain, because of the nature of our proposed framework which is based on neural networks and consequently is as simple as some consecutive matrix multiplications, it has less computational cost than conventional methods such as Dynamic Time Warping and could be used concurrently on devices such as Graphics Processing Unit and Tensor Processing Unit. [ABSTRACT FROM AUTHOR]

Convolutional neural network is powerful for general object recognition. However, its excellent performance depends largely on huge training set. Facing task like military object recognition in which image samples for training are scarce, its performance will degrade sharply. To solve this problem, a deep transfer learning method is proposed in this paper. The main idea consists of two parts: transfer learning for prior knowledge embedding and mixed layer for better feature extraction. It has been proved that the ability of feature extraction learned in large dataset is helpful to related tasks and can be transferred to a new neural network. The transfer learning process is achieved by fixing the weights of some layers and then retraining the remained layers. The key problem for deep transfer learning is which part should be transferred and which part should be retrained to adapt the network to the new task. This problem is solved by extensive experiments, and it is found that retraining the last three layers and transferring prior to the other layers can reach the best performance. Besides, we used mixed layer scheme to make use of the current information. In each mixed layer, convolution filters in different scales are combined together, helping to adapt features in different scales. By employing these two methods, the proposed method exhibits a large improvement in military object recognition under small training set. Experiments demonstrate that our method can achieve a high recognition precision, superior to many other algorithms compared. [ABSTRACT FROM AUTHOR]

Breast cancer is one of the most common female malignancies, as well as the second leading cause of mortality for women. Early detection and treatment can dramatically decrease the mortality rate. Recently, automated breast volume scanner (ABVS) has become one of the most frequently used diagnose methods for breast tumor screening because of its operator-independent and reproducible advantages. However, it is a challenging job to obtain the tumors' accurate locations and shapes by reviewing hundreds of ABVS slices. In this paper, a novel computer-aided detection (CADe) system is developed to reduce clinicians' reading time and improve the efficiency. The CADe system mainly contains three parts: tumor candidate acquisition, false-positive reduction and tumor segmentation. Firstly, a local phase-based approach is built to obtain breast tumor candidates for further recognition. Subsequently, a convolutional neural network (CNN) is applied to reduce false positives (FPs). The introduction of CNN can help to avoid complicated feature extraction as well as elevate the accuracy and efficiency. Finally, superpixel-based segmentation is used to outline the breast tumor. Here, superpixel-based local binary pattern (SLBP) is proposed to assist the segmentation, which improves the performance. The methods were evaluated on a clinical ABVS dataset whose abnormal cases were manually labeled by an experienced radiologist. The experiment results were mainly composed of two parts. At the FP reduction stage, the proposed CNN achieved 100% and 78.12% sensitivity with FPs/case of 2.16 and 0. At the segmentation stage, our SLBP obtained 82.34% true positive, 15.79% false positive and 83.59% Dice similarity. In summary, the proposed CADe system demonstrated promising potential to detect and outline breast tumors in ABVS images. [ABSTRACT FROM AUTHOR]

This paper presents a learning mechanism based on hybridization of static and dynamic learning. Realizing the detection performances offered by the state-of-the-art deep learning techniques and the competitive performances offered by the conventional static learning techniques, we propose the idea of exploitation of the concatenated (parallel) hybridization of the static and dynamic learning-based feature spaces. This is contrary to the cascaded (series) hybridization topology in which the initial feature space (provided by the conventional, static, and handcrafted feature extraction technique) is explored using deep, dynamic, and automated learning technique. Consequently, the characteristics already suppressed by the conventional representation cannot be explored by the dynamic learning technique. Instead, the proposed technique combines the conventional static and deep dynamic representation in concatenated (parallel) topology to generate an information-rich hybrid feature space. Thus, this hybrid feature space may aggregate the good characteristics of both conventional and deep representations, which are then explored using an appropriate classification technique. We also hypothesize that ensemble classification may better exploit this parallel hybrid perspective of the feature spaces. For this purpose, pyramid histogram of oriented gradients-based static learning has been incorporated in conjunction with convolution neural network-based deep learning to produce concatenated hybrid feature space. This hybrid space is then explored with various state-of-the-art ensemble classification techniques. We have considered the publicly available INRIA person and Caltech pedestrian standard image datasets to assess the performance of the proposed hybrid learning system. Furthermore, McNemar's test has been used to statistically validate the outperformance of the proposed technique over various contemporary techniques. The validated experimental results show that the employment of the proposed hybrid representation results in effective detection performance (an AUC of 0.9996 for INRIA person and 0.9985 for Caltech pedestrian datasets) as compared to the individual static and dynamic representations. [ABSTRACT FROM AUTHOR]

A numeric quantity that characterizes the whole structure of a network is called a topological index. In the studies of quantitative structure-activity relationship and quantitative structure-property relationship, the topological indices are utilized to guess the physical features related to the bioactivities and chemical reactivities in certain networks. A neural network is a computer system modeled on the nerve tissue and nervous system. The neural networks are not only studied in Neurochemistry. There are many applications of these networks in different areas of studies such as intrusion detection system, image processing, artificial intelligence, localization, medicine, chemical, and environmental sciences. In this paper, we compute the degree-based topological indices of the probabilistic neural network for the first time. At the end, a numerical comparison between all the indices is also shown with the help of the Cartesian coordinate system. [ABSTRACT FROM AUTHOR]

In this paper, an architecture based on memristors is proposed to implement image convolution computation in convolutional neural networks. This architecture could extract different features of input images when using different convolutional kernels. Bipolar memristors with threshold are employed in this work, which vary their conductance values under different voltages. Various kernels are needed to extract information of input images, while different kernels contain different weights. The memristances of bipolar memristors with threshold are convenient to be varied and kept, which make them suitable to act as the weights of kernels. The performances of the design are verified by simulation results. [ABSTRACT FROM AUTHOR]

Pyramidal neural networks (PNN) are computational systems inspired by the concept of receptive fields from the human visual system. These neural networks are designed for implicit feature extraction and have been applied in pattern recognition applications. In the original approach, the size of the receptive field within the same 2D layer is a constant parameter, while in the human visual system, the receptive field size is variable. This paper proposes a PNN with variable receptive fields determined by an evolutionary algorithm, called variable pyramidal neural network with evolutionary algorithms. We observed from experiments aiming at detecting faces in images that our approach can achieve better classification rates than the original PNN. We also observed that regions with more information (such as nose and eyes) are more emphasized by variable receptive fields. These results confirm the application of intelligent algorithms to determine adjustable receptive fields in neural networks is useful to find out relevant information for recognition task. Besides, the model is comparable to biological systems regarding the flexibility assigned to receptive fields. [ABSTRACT FROM AUTHOR]

Salient object detection model with active environment perception can substantially facilitate a wide range of applications. Conventional models primarily rely on handcrafted low-level image features or high-level features. However, these models may face great challenges in low-lighting scenario, due to the lack of well-defined features to represent saliency information in low-contrast images. In this paper, we propose a novel deep neural network framework embedded with covariance descriptor for salient object detection in low-contrast images. Several low-level features are extracted to compute their mutual covariance, which is then trained via a 7-layer convolutional neural network (CNN). The saliency map can be generated by estimating the saliency score of each region via the pre-trained CNN model. Extensive experiments have been conducted on six challenging datasets to evaluate the performance of the proposed model against ten state-of-the-art models. [ABSTRACT FROM AUTHOR]

3D point clouds are important for the reconstruction of environment. However, comparing to the artificial VR scene representation methods, 3D point clouds are more difficult to correspond to real scenes. In this paper, a method for detecting keypoints and describing scale invariant point feature of 3D point clouds is proposed. To detect, we first select keypoints as the saliency points with fast changing speed along with all principal directions of the searching area of the point cloud. The searching area is a searching keyscale which represents the unique scale size of the point cloud. Then, the descriptor is encoded based on the shape of a border or silhouette of an object to be detected or recognized. We also introduce a vote-casting-based 3D multi-scale object detection method. Experimental results based on synthetic data, real data and vote-casting scheme show that we can easily deal with the different tasks without additional information. [ABSTRACT FROM AUTHOR]

This paper presents a high-level overview of Emerging Services and Analytics in Cloud Computing virtual reality including the rationale, technologies involved, selected examples, and contributions. Examples in health care, business, climate change, and natural disaster have been demonstrated to support Emerging Services and Analytics by virtual reality, which can allow scientists to understand the complexity behind science and process thousands of data within seconds. Not only scientists with deep knowledge but also the general public can understand part of the outputs and illustrations from different types of examples and outputs from other disciplines. Businesses can achieve business agility, improvement in efficiency, monitoring the daily work, and analyzing key performance indicators up to date and in real time. We present the case of Emerging Services and Analytics by virtual reality to discuss its contributions and future directions. [ABSTRACT FROM AUTHOR]

Image contour-based feature extraction method has been applied to some fields of image recognition and virtual reality. However, image contour features are easily susceptible to factors like noise, rotation and thresholds during extraction and processing. To solve the above problem, this paper proposes a contour coding image recognition algorithm based on level set and BP neural network models. Firstly, level set model is employed to extract the contours of images. Secondly, image coding method proposed herein is used to code images horizontally, vertically and obliquely. At last, BP neural network model is trained to recognize the image codes. Validity of the proposed algorithm is verified by using a set of actual engineering part images as well as MPEG and PLANE databases. The results show that the proposed method achieves high recognition rate and requires small samples, which also exhibits good robustness to external disturbances such as noise and image scaling and rotation. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel hybrid text classification model based on deep belief network and softmax regression. To solve the sparse high-dimensional matrix computation problem of texts data, a deep belief network is introduced. After the feature extraction with DBN, softmax regression is employed to classify the text in the learned feature space. In pre-training procedures, the deep belief network and softmax regression are first trained, respectively. Then, in the fine-tuning stage, they are transformed into a coherent whole and the system parameters are optimized with Limited-memory Broyden-Fletcher-Goldfarb-Shanno algorithm. The experimental results on Reuters-21,578 and 20-Newsgroup corpus show that the proposed model can converge at fine-tuning stage and perform significantly better than the classical algorithms, such as SVM and KNN. [ABSTRACT FROM AUTHOR]

Classification of alcoholic electroencephalogram (EEG) signals is a challenging job in biomedical research for diagnosis and treatment of brain diseases of alcoholic people. The aim of this study was to introduce a robust method that can automatically identify alcoholic EEG signals based on time-frequency (T-F) image information as they convey key characteristics of EEG signals. In this paper, we propose a new hybrid method to classify automatically the alcoholic and control EEG signals. The proposed scheme is based on time-frequency images, texture image feature extraction and nonnegative least squares classifier (NNLS). In T-F analysis, the spectrogram of the short-time Fourier transform is considered. The obtained T-F images are then converted into 8-bit grayscale images. Co-occurrence of the histograms of oriented gradients (CoHOG) and Eig(Hess)-CoHOG features are extracted from T-F images. Finally, obtained features are fed into NNLS classifier as input for classify alcoholic and control EEG signals. To verify the effectiveness of the proposed approach, we replace the NNLS classifier by artificial neural networks, k-nearest neighbor, linear discriminant analysis and support vector machine classifier separately, with the same features. Experimental outcomes along with comparative evaluations with the state-of-the-art algorithms manifest that the proposed method outperforms competing algorithms. The experimental outcomes are promising, and it can be anticipated that upon its implementation in clinical practice, the proposed scheme will alleviate the onus of the physicians and expedite neurological diseases diagnosis and research. [ABSTRACT FROM AUTHOR]

Moving object recognition (MOR) is an important but challenging problem in the field of computer vision. The aim of MOR is to recognize moving objects in a given video dataset. Convolutional neural networks (CNNs) have been extensively used for image recognition and video analysis problems. Recently, a 3D-CNN, which contains 3D convolution layers, was proposed to address MOR problems by successfully extracting spatiotemporal features. In this paper, a multi-view (MV) 3D-CNN is proposed for MOR. This model combines 3D-CNNs with a well-known MV learning technique. Because multi-view learning techniques have the ability to obtain more view-related features from videos captured by different cameras, the proposed model can extract more representative features. Moreover, the model contains a special view-pooling layer that can fuse the feature information from previous layers. The proposed MV3D-CNN is applied to both real-world moving vehicle recognition and sign language recognition tasks. The experimental results show that the proposed model possesses good performance. [ABSTRACT FROM AUTHOR]

This paper presents a modified TLBO (teaching-learning-based optimization) approach for the local linear radial basis function neural network (LLRBFNN) model to classify multiple power signal disturbances. Cumulative sum average filter has been designed for localization and feature extraction of multiple power signal disturbances. The extracted features are fed as inputs to the modified TLBO-based LLRBFNN for classification. The performance of the proposed modified TLBO-based LLRBFNN model is compared with the conventional model in terms of convergence speed and classification accuracy. Also, an extreme learning machine (ELM) approach is used to optimize the performance of the proposed LLRBFNN and is compared with the TLBO method in classifying the multiple power signal disturbances. The classification results reveal that although the TLBO approach produces slightly better accuracy in comparison with the ELM approach, the latter is much faster in implementation, thus making it suitable for processing large quantum of power signal disturbance data. [ABSTRACT FROM AUTHOR]

Plant is closely related to humans. How to quickly recognize an unknown plant without related professional knowledge is a huge challenge. With the development of image processing and pattern recognition, it is available for plant recognition based on the technique of image processing. Pulse-coupled neural network is a powerful tool for image processing. It is widely applied in the field of image segmentation, image fusion, feature extraction, etc. Support vector machine is an excellent classifier, which can finish the complex task of data exploration. Based on these two techniques, a novel plant recognition method is proposed in this paper. The key feature is the entropy sequence obtained by pulse-coupled neural network. Other ancillary features can be computed directly by mathematical and morphological methods. Both key feature and ancillary features are employed to represent the unique feature of one plant. Support vector machine in our method is taken as the classifier, which can implement the multi-class classification. Experimental results show that the proposed method can finish the task of plant recognition effectively. Compared with the existing methods, our proposed method has better recognition rate. [ABSTRACT FROM AUTHOR]

Fetal heart rate helps in diagnosing the well-being and also the distress of fetal. Cardiotocograph (CTG) monitors the fetal heart activity to estimate the fetal tachogram based on the evaluation of ultrasound pulses reflected from the fetal heart. It consists in a simultaneous recording and analysis of fetal heart rate signal, uterine contraction activity and fetal movements. Generally CTG comprises more number of features. Feature selection also called as attribute selection is a process of selecting a subset of highly relevant features which is responsible for future analysis. In general, medical datasets require more number of features to predict an activity. This paper aims at identifying the relevant and ignores the redundant features, consequently reducing the number of features to assess the fetal heart rate. The features are selected by using unsupervised particle swarm optimization (PSO)-based relative reduct (US-PSO-RR) and compared with unsupervised relative reduct and principal component analysis. The proposed method is then tested by applying various classification algorithms such as single decision tree, multilayer perceptron neural network, probabilistic neural network and random forest for maximum number of classes and clustering accuracies like root mean square error, mean absolute error, Davies-Bouldin index and Xie-Beni index for minimum number of classes. Empirical results show that the US-PSO-RR feature selection technique outperforms the existing methods by producing sensitivity of 72.72 %, specificity of 97.66 %, F-measure of 74.19 % which is remarkable, and clustering results demonstrate error rate produced by US-PSO-RR is less as well. [ABSTRACT FROM AUTHOR]