Search

This article addresses the challenges encountered in underwater acoustic communication (UWAC) and presents a novel approach for chirp spread spectrum (CSS) communication. CSS is recognized for its ability to adjust to multipath and Doppler dispersion in underwater conditions, despite it usually demands a large bandwidth time product to achieve optimal performance. To address this constraint and improve data rate, the paper proposes a neural network-based receiver for spectral efficient M-ary CSS communication. M-ary communication is accomplished by transmitting chirps with different start and stop frequencies. At the receiver, a multilayer perceptron (MLP) artificial neural network and a one-dimensional convolutional neural network (1D CNN) are used for supervised classification. The neural network is trained offline using a comprehensive dataset developed by the BELLHOP ray tracing algorithm, which simulates various underwater acoustic channels. The application of VTRM pre-processing equalization aims to enhance performance. The simulation results illustrate the superior performance of the proposed receiver when compared to a conventional receiver based on a matched filter. The 16-ary chirp spread spectrum 1D CNN and MLP receivers show a gain of 6 and 4 dB, respectively, in a simulated channel after undergoing VTRM pre-processing. Furthermore, the utilization of a 16-ary 1D CNN receiver results in a noticeable 6 dB enhancement in two recorded channels. However, the MLP receiver outperforms the traditional receiver in terms of bit error rate. The article emphasizes the possibility of higher data rates and enhanced performance in underwater communication systems by employing the proposed M-ary CSS neural network-based method. [ABSTRACT FROM AUTHOR]

Speech Emotion Recognition (SER) is a challenging task, and the typical convolutional neural network (CNN) cannot well handle the speech data directly. Because CNN tends to understand local information and ignores the overall characteristics. This paper proposes a Capsule Network with Two-Way Attention MechanismTWACapsNet for short) for the SER problem. TWACapsNet accepts the spatial and spectral features as inputs, and the convolutional layer and the capsule layer are deployed to process these two types of features in two ways separately. After that, two attention mechanisms are designed to enhance the information obtained from the spatial and spectral features. Finally, the results of these two ways are combined to form the final decision. The advantage of TWACapsNet is verified by experiments on multiple SER data sets, and experimental results show that the proposed method outperforms the widely-deployed neural network models on three typical SER data sets. Furthermore, the combination of the two ways contributes to the higher and more stable performance of TWACapsNet. [ABSTRACT FROM AUTHOR]

This paper presents the results of a study on data preprocessing and modeling for predicting corrosion in water pipelines of a steel industrial plant. The use case is a cooling circuit consisting of both direct and indirect cooling. In the direct cooling circuit, water comes into direct contact with the product, whereas in the indirect one, it does not. In this study, advanced machine learning techniques, such as extreme gradient boosting and deep neural networks, have been employed for two distinct applications. Firstly, a virtual sensor was created to estimate the corrosion rate based on influencing process variables, such as pH and temperature. Secondly, a predictive tool was designed to foresee the future evolution of the corrosion rate, considering past values of both influencing variables and the corrosion rate. The results show that the most suitable algorithm for the virtual sensor approach is the dense neural network, with MAPE values of (25 ± 4)% and (11 ± 4)% for the direct and indirect circuits, respectively. In contrast, different results are obtained for the two circuits when following the predictive tool approach. For the primary circuit, the convolutional neural network yields the best results, with MAPE = 4% on the testing set, whereas for the secondary circuit, the LSTM recurrent network shows the highest prediction accuracy, with MAPE = 9%. In general, models employing temporal windows have emerged as more suitable for corrosion prediction, with model performance significantly improving with a larger dataset. [ABSTRACT FROM AUTHOR]

To reliably and accurately classify complicated "big" datasets, machine learning models must be continually improved. This research proposes straightforward yet competitive neural networks for text classification, even though graph neural networks (GNN) have reignited interest in graph-based text classification models. Convolutional neural networks (CNN), artificial neural networks (ANN), and their refined "fine-tuned" models (denoted as FT-CNN and FT-ANN) are the names given to our proposed models. The models presented in this paper demonstrate that our simple models like (CNN, ANN, FT-CNN, and FT-ANN) can perform better than more complex GNN ones such as (SGC, SSGC, and TextGCN) and are comparable to others (i.e., HyperGAT and Bert). The process of fine-tuning is also highly recommended because it improves the performance and reliability of models. The performance of our suggested models on five benchmark datasets (namely, Reuters (R8), R52, 20NewsGroup, Ohsumed, and Mr) is vividly illustrated. According to the experimental findings, on the majority of the target datasets, these models—especially those that have been fine-tuned—perform surprisingly better than SOTA approaches, including GNN-based models. [ABSTRACT FROM AUTHOR]

The rapidly evolving field of Virtual Reality (VR)-based Human–Computer Interaction (HCI) presents a significant demand for robust and accurate hand tracking solutions. Current technologies, predominantly based on single-sensing modalities, fall short in providing comprehensive information capture due to susceptibility to occlusions and environmental factors. In this paper, we introduce a novel sensor fusion approach combined with a Long Short-Term Memory (LSTM)-based algorithm for enhanced hand tracking in VR-based HCI. Our system employs six Leap Motion controllers, two RealSense depth cameras, and two Myo armbands to yield a multi-modal data capture. This rich data set is then processed using LSTM, ensuring the accurate real-time tracking of complex hand movements. The proposed system provides a powerful tool for intuitive and immersive interactions in VR environments. [ABSTRACT FROM AUTHOR]

To tune the wave energy converter (WEC) controller parameters such as damping to reduce the line force during extreme wave conditions, future knowledge of the line force is required. To achieve this, the incoming wave and system state should be predicted for a few seconds in the future. It is rather an arduous task to predict the future knowledge of waves and the system's dynamic when dealing with breaking and steep waves, and the system is subject to various nonlinear forces. The classical model-based control strategies often rely on linear assumptions to estimate the WEC dynamics for the sake of simplicity. Unlike the model-based, the data-driven approaches are free from modeling errors and the algorithms are trained over the true and noisy data to predict non-linear system behaviors. Using data-driven approaches, we are able to model nonlinear dynamics. However, new questions emerge on the accuracy of the future wave and system state predictions, and how this uncertainty propagates into the final prediction of the line force. As incorrect damping may lead to excessive line force and detrimental damage to the system, these are the knowledge gaps that need to be addressed. The main purpose of this paper is to answer these questions through a survivability strategy for wave energy converters by providing a realistic perspective on the implementation of the neural network approaches by accounting for the errors in the input data. For this purpose, a series of neural networks is designed that first predicts the surface elevation for 0.36 s ahead, i.e. corresponding to 2 s in the full-scale WEC. This future knowledge of the wave elevation is then used to predict the system state (i.e. power take-off (PTO) translator position) for the same prediction horizon based on the PTO damping. This information is then fed to a convolutional neural network (CNN) that predicts the peak line force 0.36 s ahead. This paper also evaluates the predictive capabilities of neural network (NN) models, comparing them to classical autoregressive (AR) and Kalman filter methods. While the AR model exhibits slightly higher accuracy in fixed PTO system configurations, it falters in providing real-time predictions when system parameters undergo variations. In contrast, the NN prediction model excels by establishing a robust relationship between system configuration and output signals. Especially noteworthy is its ability to outperform classical methods with minimal training on a selective set of system configurations. Then, the sensitivity of the peak line force prediction to the uncertainties in the input data from NN models and the prediction horizon is analyzed. The neural network models are trained over the experimental data subjected to extreme sea states for a point absorber wave energy converter. The results suggest that the accuracy of the surface elevation prediction has an insignificant direct effect on the peak force prediction model. However, these uncertainties are reflected in the PTO translator position prediction, and the model is considerably sensitive to the accuracy of this prediction. This sensitivity nonetheless is less notable for higher PTO damping values. • Predicting wave elevation, PTO position, and mooring force via neural networks. • Finding optimum damping to reduce mooring force in extreme wave conditions. • Predicting the mooring force is sensitive to the PTO position prediction. [ABSTRACT FROM AUTHOR]

This paper presents a method for the transparent, robust, and highly capacitive watermarking of video signals using an information mapper. The proposed architecture is based on the use of deep neural networks to embed the watermark in the luminance channel in the YUV color space. An information mapper was used to enable the transformation of a multi-bit binary signature of varying capacitance reflecting the entropy measure of the system into a watermark embedded in the signal frame. To confirm the effectiveness of the method, tests were carried out for video frames with a resolution of 256 × 256 pixels, with a watermark capacity of 4 to 16,384 bits. Transparency metrics (SSIM and PSNR) and a robustness metric—the bit error rate (BER)—were used to assess the performance of the algorithms. [ABSTRACT FROM AUTHOR]

This paper tackles the problem of training a deep convolutional neural network of both low-bitwidth weights and activations. Optimizing a low-precision network is very challenging due to the non-differentiability of the quantizer, which may result in substantial accuracy loss. To address this, we propose three practical approaches, including (i) progressive quantization; (ii) stochastic precision; and (iii) joint knowledge distillation to improve the network training. First, for progressive quantization, we propose two schemes to progressively find good local minima. Specifically, we propose to first optimize a network with quantized weights and subsequently quantize activations. This is in contrast to the traditional methods which optimize them simultaneously. Furthermore, we propose a second progressive quantization scheme which gradually decreases the bitwidth from high-precision to low-precision during training. Second, to alleviate the excessive training burden due to the multi-round training stages, we further propose a one-stage stochastic precision strategy to randomly sample and quantize sub-networks while keeping other parts in full-precision. Finally, we adopt a novel learning scheme to jointly train a full-precision model alongside the low-precision one. By doing so, the full-precision model provides hints to guide the low-precision model training and significantly improves the performance of the low-precision network. Extensive experiments on various datasets (e.g., CIFAR-100, ImageNet) show the effectiveness of the proposed methods. [ABSTRACT FROM AUTHOR]

Neural networks have shown tremendous growth in recent years to solve numerous problems. Various types of neural networks have been introduced to deal with different types of problems. However, the main goal of any neural network is to transform the non-linearly separable input data into more linearly separable abstract features using a hierarchy of layers. These layers are combinations of linear and nonlinear functions. The most popular and common non-linearity layers are activation functions (AFs), such as Logistic Sigmoid, Tanh, ReLU, ELU, Swish and Mish. In this paper, a comprehensive overview and survey is presented for AFs in neural networks for deep learning. Different classes of AFs such as Logistic Sigmoid and Tanh based, ReLU based, ELU based, and Learning based are covered. Several characteristics of AFs such as output range, monotonicity, and smoothness are also pointed out. A performance comparison is also performed among 18 state-of-the-art AFs with different networks on different types of data. The insights of AFs are presented to benefit the researchers for doing further research and practitioners to select among different choices. The code used for experimental comparison is released at: https://github.com/shivram1987/ActivationFunctions. [ABSTRACT FROM AUTHOR]

Angular momentum of light can be divided into spin angular momentum and orbital angular momentum (OAM). Due to the theoretically unlimited orthogonal states, the physical dimension of OAM provides a potential solution to boost the information capacity. The OAM multiplexing and modulation techniques have been implemented to meet the continuous growth of bandwidth requirements, resulting in the concept of OAM optical communication. However, the performances of the traditional optical OAM detection techniques degrade seriously in the practical application of OAM optical communications. Thanks to the powerful data analysis advantages, the cutting-edge machine learning (ML) algorithms have been widely used in the field of image processing, laying the technical foundation for OAM recognition. This paper reviews the recent advances on OAM optical communications that are enhanced by ML methods. More than the traditional OAM detection methods, the OAM demodulation methods based on multiple network architectures, including the support vector machine, self-organizing map, feed-forward neural network, convolutional neural network, and diffractive deep optical neural network (D2NN), have been summarized. We also discuss the development of the spiking neural network and on-chip D2NN, opening a possible way to facilitate the future ultra-low power and ultra-fast OAM demodulation technology. [ABSTRACT FROM AUTHOR]

Facial emotion recognition extracts the human emotions from the images and videos. As such, it requires an algorithm to understand and model the relationships between faces and facial expressions and to recognize human emotions. Recently, deep learning models are utilized to improve the performance of facial emotion recognition. However, the deep learning models suffer from the overfitting issue. Moreover, deep learning models perform poorly for images which have poor visibility and noise. Therefore, in this paper, an efficient deep learning-based facial emotion recognition model is proposed. Initially, contrast-limited adaptive histogram equalization (CLAHE) is applied to improve the visibility of input images. Thereafter, a modified joint trilateral filter is applied to the obtained enhanced images to remove the impact of impulsive noise. Finally, an efficient deep convolutional neural network is designed. Adam optimizer is also utilized to optimize the cost function of deep convolutional neural networks. Experiments are conducted by using the benchmark dataset and competitive human emotion recognition models. Comparative analysis demonstrates that the proposed facial emotion recognition model performs considerably better compared to the competitive models [ABSTRACT FROM AUTHOR]

Deep convolutional neural networks (DCNNs) are currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose $\Pi$ Π -Nets, a new class of function approximators based on polynomial expansions. $\Pi$ Π -Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. The unknown parameters, which are naturally represented by high-order tensors, are estimated through a collective tensor factorization with factors sharing. We introduce three tensor decompositions that significantly reduce the number of parameters and show how they can be efficiently implemented by hierarchical neural networks. We empirically demonstrate that $\Pi$ Π -Nets are very expressive and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, $\Pi$ Π -Nets produce state-of-the-art results in three challenging tasks, i.e., image generation, face verification and 3D mesh representation learning. The source code is available at https://github.com/grigorisg9gr/polynomial_nets. [ABSTRACT FROM AUTHOR]

While thunderstorms can pose severe risks to property and life, forecasting remains challenging, even at short lead times, as these often arise in meta-stable atmospheric conditions. In this paper, we examine the question of how well we could perform short-term (up to 180 min) forecasts using exclusively multi-spectral satellite images and past lighting events as data. We employ representation learning based on deep convolutional neural networks in an "end-to-end" fashion. Here, a crucial problem is handling the imbalance of the positive and negative classes appropriately in order to be able to obtain predictive results (which is not addressed by many previous machine-learning-based approaches). The resulting network outperforms previous methods based on physically based features and optical flow methods (similar to operational prediction models) and generalizes across different years. A closer examination of the classifier performance over time and under masking of input data indicates that the learned model actually draws most information from structures in the visible spectrum, with infrared imaging sustaining some classification performance during the night. [ABSTRACT FROM AUTHOR]

Image super-resolution (SR) task aims to recover high-resolution (HR) images from degraded low-resolution (LR) images, which has achieved great progress due to the recent advances of deep neural networks. Due to severe information loss of the LR images, it is more challenging to reconstruct high quality HR images at large scale factors, i. e., higher than $4\times $. Traditional reference image based SR methods usually perform patch matching to locate detailed texture from HR reference images which could provide fine details from similar image contents. But it suffers from difficulties in achieving good matching in the largely downscaled image space or feature space due to the ill-posed nature between LR and HR mapping. In this paper, we tackle this problem by exploiting fine details contained in reference HR images. Inspired by vector quantization (VQ), we propose a simple yet effective auto-encoder convolutional neural network (CNN) module to learn discrete representations of images. Furthermore, we propose to progressively learn pairs of cross-scale discrete feature representations using paired LR and HR reference images. The coarser scale of the discrete representation is responsible for encoding the global image structure while the paired finer scale of the discrete representation takes charge of capturing missing details in the finer image scale. During inference, continuous features of the test LR image are used as queries to retrieve finer scale discrete representations (value) by searching the nearest coarser scale discrete representations (key). Then, the queries and retrieved values are combined to progressively recover the HR image. Experimental results indicate that when compared with the state-of-the-art image SR models, the proposed method can achieve advanced performance in terms of both objective quality and subjective quality. The code will be available on URL: https://github.com/sunwj/refsr. [ABSTRACT FROM AUTHOR]

The paper proposes a novel block-wise prediction paradigm based on Convolutional Neural Networks (CNNs) for lossless video coding. A deep neural network model which follows a multi-resolution design is employed for block-wise prediction. Several contributions are proposed to improve neural network training. A first contribution proposes a novel loss function formulation for an efficient network training based on a new approach for patch selection. Another contribution consists in replacing all HEVC-based angular intra-prediction modes with a CNN-based intra-prediction method, where each angular prediction mode is complemented by a CNN-based prediction mode using a specifically trained model. Another contribution consists in an efficient adaptation of the CNN-based intra-prediction residual for lossless video coding. Experimental results on standard test sequences show that the proposed coding system outperforms the HEVC standard with an average bitrate improvement of around 5%. To our knowledge, the paper is the first to replace all the traditional HEVC-based angular intra-prediction modes with an intra-prediction method based on modern Machine Learning techniques for lossless video coding applications. [ABSTRACT FROM AUTHOR]

This paper presents an approach to optimize a Convolutional Neural Network using the Fuzzy Gravitational Search Algorithm. The optimized parameters are the number of images per block that are used in the training phase, the number of filters and the filter size of the convolutional layer. The reason for optimizing these parameters is because they have a great impact on performance of the Convolutional Neural Networks. The neural network model presented in this work can be applied for any image recognition or classification applications; nevertheless, in this paper, the experiments are performed in the ORL and Cropped Yale databases. The results are compared with other neural networks, such as modular and monolithic neural networks. In addition, the experiments were performed manually, and the results were obtained (when the neural network is not optimized), and comparison was made with the optimized results to validate the advantage of using the Fuzzy Gravitational Search Algorithm. [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]

In this paper, a novel perceptual image hashing scheme based on convolutional neural network (CNN) with multiple constraints is proposed, in which our deep hashing network learns the process of features extraction automatically according to the training target and then generates the final hash sequence. The combination of convolutional and pooling layers is to reduce the size of input image while deepening the channels. Then, we construct two pairs of constraints and integrate them into an overall constraint function through a strategy of weight allocation. In order to guarantee the robustness and discrimination of deep hashing network simultaneously, a new training method is developed to adjust the training set structure dynamically according to the changes of constraint values. Experimental results show that the proposed deep hashing network can achieve a satisfactory balance between perceptual robustnzess and discrimination while maintaining security. Based on the large-scale test set, receiver operating characteristic (ROC) curves, ${F}_{1}$ scores and equal error rate (EER) demonstrate the superiority of our scheme in terms of content authentication compared with some state-of-the-art schemes. [ABSTRACT FROM AUTHOR]

In this paper, we present a method to mine object-part patterns from conv-layers of a pre-trained convolutional neural network (CNN). The mined object-part patterns are organized by an And-Or graph (AOG). This interpretable AOG representation consists of a four-layer semantic hierarchy, i.e., semantic parts, part templates, latent patterns, and neural units. The AOG associates each object part with certain neural units in feature maps of conv-layers. The AOG is constructed with very few annotations (e.g., 3–20) of object parts. We develop a question-answering (QA) method that uses active human-computer communications to mine patterns from a pre-trained CNN, in order to explain features in conv-layers incrementally. During the learning process, our QA method uses the current AOG for part localization. The QA method actively identifies objects, whose feature maps cannot be explained by the AOG. Then, our method asks people to annotate parts on the unexplained objects, and uses answers to discover CNN patterns corresponding to newly labeled parts. In this way, our method gradually grows new branches and refines existing branches on the AOG to semanticize CNN representations. In experiments, our method exhibited a high learning efficiency. Our method used about $1/6$ 1 / 6 – $1/3$ 1 / 3 of the part annotations for training, but achieved similar or better part-localization performance than fast-RCNN methods. [ABSTRACT FROM AUTHOR]

This paper proposes a generic method to learn interpretable convolutional filters in a deep convolutional neural network (CNN) for object classification, where each interpretable filter encodes features of a specific object part. Our method does not require additional annotations of object parts or textures for supervision. Instead, we use the same training data as traditional CNNs. Our method automatically assigns each interpretable filter in a high conv-layer with an object part of a certain category during the learning process. Such explicit knowledge representations in conv-layers of the CNN help people clarify the logic encoded in the CNN, i.e., answering what patterns the CNN extracts from an input image and uses for prediction. We have tested our method using different benchmark CNNs with various architectures to demonstrate the broad applicability of our method. Experiments have shown that our interpretable filters are much more semantically meaningful than traditional filters. [ABSTRACT FROM AUTHOR]

The key role of spectrum sensing in cognitive radios attracted substantial research attention to improve the performance of detectors. We consider a general model for the receiver noise with the potential of generalization towards modeling noise in various environments. This general noise model describes accurately noise characteristics ranging from the Gaussian noise to the severe impulsive noise. However, many previous studies are based on the ideal Gaussian noise, and they cannot capture the non-Gaussian models. To provide a robust detector against different behaviors of the noise in various environments, we employ a convolutional neural network (CNN) compatible with different noise models. The proposed CNN detector is data-driven, and due to its single-dimensional input layer, it is consistent with the received signal and requires no pre-processing. The likelihood ratio test (LRT), the Wald, and the Rao tests for this problem are derived to enrich the paper with comparative evaluations of the proposed CNN and conventional model-based approaches and other neural networks. Although various simulated scenarios substantiate the general superiority and robustness of the CNN-based against impulsive noise and mismatch of parameters, it requires higher computational complexity than other discussed detectors. [ABSTRACT FROM AUTHOR]

Today, a significant content of the system on chips (SOCs) is dedicated to static random access memory (SRAM) cells. Due to the stress during SRAM operation, MOSFET aging is becoming an increasingly serious problem. The manufacturing variability of the MOSFETs causes further differences in the aging of various devices. This paper presents a simulation methodology using a deep neural network (DNN) to predict the performance of SRAM cells, taking into account the impacts of fabrication variations, temperature, and aging. A dataset of four input features and six output criteria is utilized, which are derived from power consumption, power supply transient current, signal rise time, read static noise margin (RSNM), and butterfly and N-curve. The accuracy of the long short-term memory (LSTM) model in predicting the performance of SRAM cells outperformed the other models such as multilayer perceptron (MLP), one-dimensional convolutional neural network (1D-CNN), gated recurrent unit (GRU), and recurrent neural network (RNN) by 1.85 % to 1.11 %. For instance, by processing the supply current data, the LSTM model achieved 97.41 %, 97.96 %, and 97.41 % accuracy in predicting the RMS power consumption, the signal rise time, and the average power supply current of the SRAM cell, respectively. • DNN successfully predicts the aging and process variation impact on SRAM performance. • Supply current illustrates superior results in performance prediction. • LSTM DNN model outperforms MLP, RNN, 1D-CNN, and GRU, exhibiting sufficient accuracy. [ABSTRACT FROM AUTHOR]

For the emerging mobility-on-demand services, it is of great significance to predict passenger demands based on historical mobility trips towards better vehicle distribution. Prior works have focused on predicting next-step passenger demands at selected locations or hotspots. However, we argue that multi-step citywide passenger demands encapsulate both time-varying demand trends and global statuses, and hence are more beneficial to avoiding demand-service mismatching and developing effective vehicle distribution/scheduling strategies. Furthermore, we find that adaptations of single-step methods are unable to achieve robust prediction with high accuracy for further steps. In this paper, we propose an end-to-end deep neural network model to the prediction task. We employ an encoder-decoder framework based on convolutional and ConvLSTM units to identify complex features that capture spatiotemporal influence and pickup-dropoff interactions on citywide passenger demands. We introduce a multi-level attention model (global attention and temporal attention) to emphasize the effects of latent citywide mobility regularities and capture relevant temporal dependencies. We evaluate our proposed method using real-world mobility trips (taxis and bikes) and the experimental results show that our method achieves higher prediction accuracy than the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Misclassification is a critical problem in many machine learning applications. Since even the classifier models with high accuracy (e.g., > 95%) still introduce some misclassification error, it may not be possible to rely on the output of a classifier. In this paper, we introduce trustable learning, which prompts the learning model to yield only the true output, thus avoiding misclassifications. Whenever the model cannot decide the output accurately, the learning model should indicate that there could be a misclassification error if it is forced to classify, and hence, it should reject to make a decision or defer it to a human expert. Therefore, we develop a methodology for trustable learning and apply it to artificial neural networks and show that it is possible to develop a classifier with 0% misclassification error. We propose a novel neural network architecture named WisdomNet that could provide zero prediction error by introducing an additional neuron named as conjugate neuron that would indicate whether the network is able to classify the data correctly or not. The WisdomNet architecture can be applied to any previously built model, and we have evaluated WisdomNet with several network architectures such as multilayer perceptron, convolutional neural network, and deep network on different data sets. The results show that the WisdomNet is able to reduce the classification error rate to 0%, while labeling the data is difficult to classify as 'reject' at a low percentage of within around 10%. [ABSTRACT FROM AUTHOR]

With only coarse labels, weakly supervised learning typically uses top-down attention maps generated by back-propagating gradients as priors for tasks such as object localization and semantic segmentation. While these attention maps are intuitive and informative explanations of deep neural network, there is no effective mechanism to manipulate the network attention during learning process. In this paper, we address three shortcomings of previous approaches in modeling such attention maps in one common framework. First, we make attention maps a natural and explicit component in the training pipeline such that they are end-to-end trainable. Moreover, we provide self-guidance directly on these maps by exploring supervision from the network itself to improve them towards specific target tasks. Lastly, we proposed a design to seamlessly bridge the gap between using weak and extra supervision if available. Despite its simplicity, experiments on the semantic segmentation task demonstrate the effectiveness of our methods. Besides, the proposed framework provides a way not only explaining the focus of the learner but also feeding back with direct guidance towards specific tasks. Under mild assumptions our method can also be understood as a plug-in to existing convolutional neural networks to improve their generalization performance. [ABSTRACT FROM AUTHOR]

This paper presents an efficient OCR system for the recognition of offline Pashto isolated characters. The lack of an appropriate dataset makes it challenging to match against a reference and perform recognition. This research work addresses this problem by developing a medium-size database that comprises 4488 samples of handwritten Pashto character; that can be further used for experimental purposes. In the proposed OCR system the recognition task is performed using convolution neural network. The performance analysis of the proposed OCR system is validated by comparing its results with artificial neural network and support vector machine based on zoning feature extraction technique. The results of the proposed experiments shows an accuracy of 56% for the support vector machine, 78% for artificial neural network, and 80.7% for the proposed OCR system. The high recognition rate shows that the OCR system based on convolution neural network performs best among the used techniques. [ABSTRACT FROM AUTHOR]

Rear-end collision prediction has gained an increasing attention for safety improvement in smart cities. It is urgent to design efficient warning strategies for rear-end collisions which is one of the main causes of traffic accidents. The existing researches have been conducted to predict the collisions. Learning-based methods are proposed to solve this complicated issue, which the traditional methods are difficult to solve. However, due to some limitations in terms of the feature extraction and prediction performance, back-propagation learning methods are facing challenge. In this paper, we proposed a novel Rear-end Collision Prediction Mechanism with deep learning method (RCPM), in which a convolutional neural network model is established. In RCPM, the dataset is smoothed and expanded based on genetic theory to alleviate the class imbalance problem. The preprocessed dataset is divided into training and testing sets as the input to train our convolutional neural network model. The experimental results show that compared with the Honda, Berkeley and multi-layer perception neural-network-based algorithms, RCPM effectively improves performance to predict rear-end collisions. [ABSTRACT FROM AUTHOR]

When the elevation of target is smaller than a beamwidth, the complex multipath signals will distort the feature of direct signal reflected from target. The elevation of target can hardly be estimated accurately. Hence, in this paper, we proposed three kinds of neural networks models including deep neural network (DNN), 1-D convolutional neural network (1-D CNN) and 2-D convolutional neural network (2-D CNN) and their optimization method to mitigate phase distortion caused by multipath signals and enhance the phase feature of direct signal. The direction of arrival (DOA) estimation accuracy of physics-driven methods including digital beamforming (DBF) and multiple signal classification (MUSIC) is effectively improved. Concretely, we analyze the origins of error of DOA estimation in multipath environment and discuss the importance of phase feature to DOA estimation. A complete framework of feature-to-feature phase enhancement is built for DOA estimation in radar systems. The results of experiments with real data collected from a very high frequency (VHF) radar demonstrate the superior DOA estimation performance of proposed feature-to-feature learning methods with respect to other state-of-the-art methods including physics-driven methods and existing data-driven methods. [ABSTRACT FROM AUTHOR]

Convolutional neural networks are the most widely used type of neural networks in applications. In mathematical analysis, however, mostly fully-connected networks are studied. In this paper, we establish a connection between both network architectures. Using this connection, we show that all upper and lower bounds concerning approximation rates of fully-connected neural networks for functions ƒ ∈ C--for an arbitrary function class C--translate to essentially the same bounds concerning approximation rates of convolutional neural networks for functions ƒ ∈ Cequi, with the class Cequi consisting of all translation equivariant functions whose first coordinate belongs to C. All presented results consider exclusively the case of convolutional neural networks without any pooling operation and with circular convolutions, i.e., not based on zero-padding. [ABSTRACT FROM AUTHOR]

Diabetic patients are quite hesitant in engaging in normal physiological activities due to difficulties associated with diabetes management. Over the last few decades, there have been advancements in the computational power of embedded systems and glucose sensing technologies. These advancements have attracted the attention of researchers around the globe developing automatic insulin delivery systems. In this paper, a method of closed-loop control of diabetes based on neural networks is proposed. These neural networks are used for making predictions based on the clinical data of a patient. A neural network feedback controller is also designed to provide a glycemic response by regulating the insulin infusion rate. An activity recognition model based on convolutional neural networks is also proposed for predicting the patient's current physical activity. Predictions from this model are transformed into a six-level code and are fed as input to the neural network glucose prediction model. Experimental results of the proposed system show good performance in keeping blood glucose levels in the nondiabetic range. [ABSTRACT FROM AUTHOR]

Various microreactor designs under development aim at fulfilling electricity and heat production requirements affordably and reliably in a variety of applications, like power generation for remote communities and military bases. The fission battery deployment strategy for microreactors provides battery-like simplicity for its installation and operation. A successful fission battery must be cost competitive, have unit standardization, enable flexible installation, possess high reliability, and feature unattended operations. Due to the design and operational features of microreactors, successful unattended operations of the system must be enabled through an on-line monitoring (OLM) system that can effectively and reliably detect and diagnose any fault in the reactor. This research introduces the concept and approach of an intelligent Fault Detection and Diagnosis Monitoring System (FDDMS) that can provide power transient dependent fault detection and diagnosis to fit within a future autonomous control framework. Specifically, this paper develops the first FDDMS module, the Power Transient Module, to reliably identify reactor operations as steady state, ramping up in power, or ramping down in power using data from a broad-scope simulator. Knowledge of the current power transient in the reactor is paramount for effective fault detection and diagnosis in any operational regime. This paper explores various data-driven methods to accurately evaluate the power transients: 1) Principal Component Analysis and Support Vector Machines, 2) Deep Neural Networks, and 3) Convolutional Neural Networks. To prepare the sensor data collected from the simulator for training the data-driven models, several preprocessing techniques were tested to optimize for evaluation accuracy. The final model selected for the Power Transient Module produced perfect evaluation results for several power transients, including in previously unseen regimes. Foundation for further module development of the FDDMS was established. Future works will see the implementation of the remaining FDDMS architecture. • Fault Detection and Diagnosis Monitoring System introduced to provide basis for reactor operations with minimal to no personnel on-site. • Several data-driven models tested and compared to accurately evaluate different power transients present in reactor operations. • Several data transformation methods explored to create 2D representations from sensor data to illustrate multi-modal temporal relationships. [ABSTRACT FROM AUTHOR]

Convolutional Neural Networks (CNNs) have been widely used in machine learning tasks. While delivering state-of-the-art accuracy, CNNs are known as both compute- and memory-intensive. This paper presents the SqueezeFlow accelerator architecture that exploits sparsity of CNN models for increased efficiency. Unlike prior accelerators that trade complexity for flexibility, SqueezeFlow exploits concise convolution rules to benefit from the reduction of computation and memory accesses as well as the acceleration of existing dense architectures without intrusive PE modifications. Specifically, SqueezeFlow employs a PT-OS-sparse dataflow that removes the ineffective computations while maintaining the regularity of CNN computations. We present a full design down to the layout at 65 $nm$nm, with an area of 4.80 $\mathrm{mm^2}$ mm 2 and power of 536.09 $\mathrm{mW}$ mW. The experiments show that SqueezeFlow achieves a speedup of $2.9\times$2.9× on VGG16 compared to the dense architectures, with an area and power overhead of only 8.8 and 15.3 percent, respectively. On three representative sparse CNNs, SqueezeFlow improves the performance and energy efficiency by $1.8\times$1.8× and $1.5\times$1.5× over the state-of-the-art sparse accelerators. [ABSTRACT FROM AUTHOR]

Frequency filtering is widely used in routine processing of seismic data to improve the signal-to-noise ratio (SNR) of recorded signals and by doing so to improve subsequent analyses. In this paper, we develop a new denoising/decomposition method, DeepDenoiser, based on a deep neural network. This network is able to simultaneously learn a sparse representation of data in the time–frequency domain and a non-linear function that maps this representation into masks that decompose input data into a signal of interest and noise (defined as any non-seismic signal). We show that DeepDenoiser achieves impressive denoising of seismic signals even when the signal and noise share a common frequency band. Because the noise statistics are automatically learned from data and require no assumptions, our method properly handles white noise, a variety of colored noise, and non-earthquake signals. DeepDenoiser can significantly improve the SNR with minimal changes in the waveform shape of interest, even in the presence of high noise levels. We demonstrate the effect of our method on improving earthquake detection. There are clear applications of DeepDenoiser to seismic imaging, micro-seismic monitoring, and preprocessing of ambient noise data. We also note that the potential applications of our approach are not limited to these applications or even to earthquake data and that our approach can be adapted to diverse signals and applications in other settings. [ABSTRACT FROM AUTHOR]

Residual networks (ResNets) have introduced a milestone for the deep learning community due to their outstanding performance in diverse applications. They enable efficient training of increasingly deep networks, reducing the training difficulty and error. The main intuition behind them is that, instead of mapping the input information, they are mapping a residual part of it. Since the original work, a lot of extensions have been proposed to improve information mapping. In this paper, a novel extension of the residual block is proposed inspired by linear dynamical systems (LDSs), called LDS-ResNet. Specifically, a new module is presented that improves mapping of residual information by transforming it in a hidden state and then mapping it back to the desired feature space using convolutional layers. The proposed module is utilized to construct multi-branch residual blocks for convolutional neural networks. An exploration of possible architectural choices is presented and evaluated. Experimental results show that LDS-ResNet outperforms the original ResNet in image classification and object detection tasks on public datasets such as CIFAR-10/100, ImageNet, VOC, and MOT2017. Moreover, its performance boost is complementary to other extensions of the original network such as pre-activation and bottleneck, as well as stochastic training and Squeeze-Excitation. [ABSTRACT FROM AUTHOR]

The main challenges of visual object tracking arise from the arbitrary appearance of the objects that need to be tracked. Most existing algorithms try to solve this problem by training a new model to regenerate or classify each tracked object. As a result, the model needs to be initialized and retrained for each new object. In this paper, we propose to track different objects in an object-independent approach with a novel two-flow convolutional neural network (YCNN). The YCNN takes two inputs (one is an object image patch, the other is a larger searching image patch), then outputs a response map which predicts how likely and where the object would appear in the search patch. Unlike the object-specific approaches, the YCNN is actually trained to measure the similarity between the two image patches. Thus, this model will not be limited to any specific object. Furthermore, the network is end-to-end trained to extract both shallow and deep dedicated convolutional features for visual tracking. And once properly trained, the YCNN can be used to track all kinds of objects without further training and updating. As a result, our algorithm is able to run at a very high speed of 45 frames-per-second. The effectiveness of the proposed algorithm can also be proved by the experiments on two popular data sets: OTB-100 and VOT-2014. [ABSTRACT FROM AUTHOR]

In this paper, we address the challenging task of scene segmentation. In order to capture the rich contextual dependencies over image regions, we propose Directed Acyclic Graph-Recurrent Neural Networks (DAG-RNN) to perform context aggregation over locally connected feature maps. More specifically, DAG-RNN is placed on top of pre-trained CNN (feature extractor) to embed context into local features so that their representative capability can be enhanced. In comparison with plain CNN (as in Fully Convolutional Networks-FCN), DAG-RNN is empirically found to be significantly more effective at aggregating context. Therefore, DAG-RNN demonstrates noticeably performance superiority over FCNs on scene segmentation. Besides, DAG-RNN entails dramatically less parameters as well as demands fewer computation operations, which makes DAG-RNN more favorable to be potentially applied on resource-constrained embedded devices. Meanwhile, the class occurrence frequencies are extremely imbalanced in scene segmentation, so we propose a novel class-weighted loss to train the segmentation network. The loss distributes reasonably higher attention weights to infrequent classes during network training, which is essential to boost their parsing performance. We evaluate our segmentation network on three challenging public scene segmentation benchmarks: Sift Flow, Pascal Context and COCO Stuff. On top of them, we achieve very impressive segmentation performance. [ABSTRACT FROM PUBLISHER]

In this paper, we seek to improve deep neural networks by generalizing the pooling operations that play a central role in the current architectures. We pursue a careful exploration of approaches to allow pooling to learn and to adapt to complex and variable patterns. The two primary directions lie in: (1) learning a pooling function via (two strategies of) combining of max and average pooling, and (2) learning a pooling function in the form of a tree-structured fusion of pooling filters that are themselves learned. In our experiments every generalized pooling operation we explore improves performance when used in place of average or max pooling. We experimentally demonstrate that the proposed pooling operations provide a boost in invariance properties relative to conventional pooling and set the state of the art on several widely adopted benchmark datasets. These benefits come with only a light increase in computational overhead during training (ranging from additional 5 to 15 percent in time complexity) and a very modest increase in the number of model parameters (e.g., additional 1, 9, and 27 parameters for mixed, gated, and 2-level tree pooling operators, respectively). To gain more insights about our proposed pooling methods, we also visualize the learned pooling masks and the embeddings of the internal feature responses for different pooling operations. Our proposed pooling operations are easy to implement and can be applied within various deep neural network architectures. [ABSTRACT FROM PUBLISHER]

It is a feasible and promising way to utilize deep neural networks to learn and extract valuable features from synthetic aperture radar (SAR) images for SAR automatic target recognition (ATR). However, it is too difficult to effectively train the deep neural networks with limited raw SAR images. In this paper, we propose a new approach to do SAR ATR, in which a multiview deep learning framework was employed. Based on the multiview SAR ATR pattern, we first present a flexible mean to generate adequate multiview SAR data, which can guarantee a large amount of inputs for network training without needing many raw SAR images. Then, a unique deep convolutional neural network containing a parallel network topology with multiple inputs is adopted. The features of input SAR images from different views will be learned by the proposed network layer by layer; meanwhile, the learned features from the distinct views are fused in different layers progressively. Therefore, the proposed framework is able to achieve a superior recognition performance, and requires only a small number of raw SAR images for network training samples generation. Experimental results have shown the superiority of the proposed framework based on the Moving and Stationary Target Acquisition and Recognition data set. [ABSTRACT FROM AUTHOR]

Crowd counting is a challenging task, mainly due to the severe occlusions among dense crowds. This paper aims to take a broader view to address crowd counting from the perspective of semantic modeling. In essence, crowd counting is a task of pedestrian semantic analysis involving three key factors: pedestrians, heads, and their context structure. The information of different body parts is an important cue to help us judge whether there exists a person at a certain position. Existing methods usually perform crowd counting from the perspective of directly modeling the visual properties of either the whole body or the heads only, without explicitly capturing the composite body-part semantic structure information that is crucial for crowd counting. In our approach, we first formulate the key factors of crowd counting as semantic scene models. Then, we convert the crowd counting problem into a multi-task learning problem, such that the semantic scene models are turned into different sub-tasks. Finally, the deep convolutional neural networks are used to learn the sub-tasks in a unified scheme. Our approach encodes the semantic nature of crowd counting and provides a novel solution in terms of pedestrian semantic analysis. In experiments, our approach outperforms the state-of-the-art methods on four benchmark crowd counting data sets. The semantic structure information is demonstrated to be an effective cue in scene of crowd counting. [ABSTRACT FROM PUBLISHER]

This paper presents a simple yet effective supervised deep hash approach that constructs binary hash codes from labeled data for large-scale image search. We assume that the semantic labels are governed by several latent attributes with each attribute on or off, and classification relies on these attributes. Based on this assumption, our approach, dubbed supervised semantics-preserving deep hashing (SSDH), constructs hash functions as a latent layer in a deep network and the binary codes are learned by minimizing an objective function defined over classification error and other desirable hash codes properties. With this design, SSDH has a nice characteristic that classification and retrieval are unified in a single learning model. Moreover, SSDH performs joint learning of image representations, hash codes, and classification in a point-wised manner, and thus is scalable to large-scale datasets. SSDH is simple and can be realized by a slight enhancement of an existing deep architecture for classification; yet it is effective and outperforms other hashing approaches on several benchmarks and large datasets. Compared with state-of-the-art approaches, SSDH achieves higher retrieval accuracy, while the classification performance is not sacrificed. [ABSTRACT FROM PUBLISHER]

Structural damage detection has been an interdisciplinary area of interest for various engineering fields. While the available damage detection methods have been in the process of adapting machine learning concepts, most machine learning based methods extract “hand-crafted” features which are fixed and manually selected in advance. Their performance varies significantly among various patterns of data depending on the particular structure under analysis. Convolutional neural networks (CNNs), on the other hand, can fuse and simultaneously optimize two major sets of an assessment task (feature extraction and classification) into a single learning block during the training phase. This ability not only provides an improved classification performance but also yields a superior computational efficiency. 1D CNNs have recently achieved state-of-the-art performance in vibration-based structural damage detection; however, it has been reported that the training of the CNNs requires significant amount of measurements especially in large structures. In order to overcome this limitation, this paper presents an enhanced CNN-based approach that requires only two measurement sets regardless of the size of the structure. This approach is verified using the experimental data of the Phase II benchmark problem of structural health monitoring which had been introduced by IASC-ASCE Structural Health Monitoring Task Group. As a result, it is shown that the enhanced CNN-based approach successfully estimated the actual amount of damage for the nine damage scenarios of the benchmark study. [ABSTRACT FROM AUTHOR]

Following the great success of deep convolutional neural networks (CNNs) in computer vision, this paper proposes a complex-valued CNN (CV-CNN) specifically for synthetic aperture radar (SAR) image interpretation. It utilizes both amplitude and phase information of complex SAR imagery. All elements of CNN including input-output layer, convolution layer, activation function, and pooling layer are extended to the complex domain. Moreover, a complex backpropagation algorithm based on stochastic gradient descent is derived for CV-CNN training. The proposed CV-CNN is then tested on the typical polarimetric SAR image classification task which classifies each pixel into known terrain types via supervised training. Experiments with the benchmark data sets of Flevoland and Oberpfaffenhofen show that the classification error can be further reduced if employing CV-CNN instead of conventional real-valued CNN with the same degrees of freedom. The performance of CV-CNN is comparable to that of existing state-of-the-art methods in terms of overall classification accuracy. [ABSTRACT FROM PUBLISHER]

As the acquisition of very high resolution (VHR) satellite images becomes easier owing to technological advancements, ever more stringent requirements are being imposed on automatic image interpretation. Moreover, per-pixel classification has become the focus of research interests in this regard. However, the efficient and effective processing and the interpretation of VHR satellite images remain a critical task. Convolutional neural networks (CNNs) have recently been applied to VHR satellite images with considerable success. However, the prevalent CNN models accept input data of fixed sizes and train the classifier using features extracted directly from the convolutional stages or the fully connected layers, which cannot yield pixel-to-pixel classifications. Moreover, training a CNN model requires large amounts of labeled reference data. These are challenging to obtain because per-pixel labeled VHR satellite images are not open access. In this paper, we propose a framework called the unsupervised-restricted deconvolutional neural network (URDNN). It can solve these problems by learning an end-to-end and pixel-to-pixel classification and handling a VHR classification using a fully convolutional network and a small number of labeled pixels. In URDNN, supervised learning is always under the restriction of unsupervised learning, which serves to constrain and aid supervised training in learning more generalized and abstract feature. To some degree, it will try to reduce the problems of overfitting and undertraining, which arise from the scarcity of labeled training data, and to gain better classification results using fewer training samples. It improves the generality of the classification model. We tested the proposed URDNN on images from the Geoeye and Quickbird sensors and obtained satisfactory results with the highest overall accuracy (OA) achieved as 0.977 and 0.989, respectively. Experiments showed that the combined effects of additional kernels and stages may have produced better results, and two-stage URDNN consistently produced a more stable result. We compared URDNN with four methods and found that with a small ratio of selected labeled data items, it yielded the highest and most stable results, whereas the accuracy values of the other methods quickly decreased. For some categories with fewer training pixels, accuracy for categories from other methods was considerably worse than that in URDNN, with the largest difference reaching almost 10%. Hence, the proposed URDNN can successfully handle the VHR image classification using a small number of labeled pixels. Furthermore, it is more effective than state-of-the-art methods. [ABSTRACT FROM PUBLISHER]

This paper addresses the problem of generating controllers for automatic bucket filling using data collected during normal operations by professional operators on worksites. We introduce methods to annotate different phases of bucket fillings from long time series data and synthesize five machine learning models. The methods learn a policy that maps input signals to control commands using imitation learning from the training data. We use four different datasets as training data and first compare the model's accuracy on predicting operator commands. We then deploy and evaluate the efficacy and robustness of the policies on a real machine. The experiments shows that the Multilayer Perceptron and the Convolutional Neural Network models had the best overall performance and robustness and could achieve human level performance. Furthermore, models with high prediction accuracy are not necessarily suitable for feedback control, and using data from real worksites increased the controller's robustness. • Automatic labeling of wheel loader bucket fillings from time series data. • Evaluating machine learning models with different datasets collected from worksites. • Human level performance for selected synthesized machine learning controllers. [ABSTRACT FROM AUTHOR]

In this paper, we propose deformable deep convolutional neural networks for generic object detection. This new deep learning object detection framework has innovations in multiple aspects. In the proposed new deep architecture, a new deformation constrained pooling (def-pooling) layer models the deformation of object parts with geometric constraint and penalty. A new pre-training strategy is proposed to learn feature representations more suitable for the object detection task and with good generalization capability. By changing the net structures, training strategies, adding and removing some key components in the detection pipeline, a set of models with large diversity are obtained, which significantly improves the effectiveness of model averaging. The proposed approach improves the mean averaged precision obtained by RCNN [1] , which was the state-of-the-art, from $31$ to $50.3$ percent on the ILSVRC2014 detection test set. It also outperforms the winner of ILSVRC2014, GoogLeNet, by 6.1 percent. Detailed component-wise analysis is also provided through extensive experimental evaluation, which provides a global view for people to understand the deep learning object detection pipeline. [ABSTRACT FROM PUBLISHER]

Before the big data era, scene recognition was often approached with two-step inference using localized intermediate representations (objects, topics, and so on). One of such approaches is the semantic manifold (SM), in which patches and images are modeled as points in a semantic probability simplex. Patch models are learned resorting to weak supervision via image labels, which leads to the problem of scene categories co-occurring in this semantic space. Fortunately, each category has its own co-occurrence patterns that are consistent across the images in that category. Thus, discovering and modeling these patterns are critical to improve the recognition performance in this representation. Since the emergence of large data sets, such as ImageNet and Places, these approaches have been relegated in favor of the much more powerful convolutional neural networks (CNNs), which can automatically learn multi-layered representations from the data. In this paper, we address many limitations of the original SM approach and related works. We propose discriminative patch representations using neural networks and further propose a hybrid architecture in which the semantic manifold is built on top of multiscale CNNs. Both representations can be computed significantly faster than the Gaussian mixture models of the original SM. To combine multiple scales, spatial relations, and multiple features, we formulate rich context models using Markov random fields. To solve the optimization problem, we analyze global and local approaches, where a top–down hierarchical algorithm has the best performance. Experimental results show that exploiting different types of contextual relations jointly consistently improves the recognition accuracy. [ABSTRACT FROM AUTHOR]

Style transfer is a process of migrating a style from a given image to the content of another, synthesizing a new image, which is an artistic mixture of the two. Recent work on this problem adopting convolutional neural-networks (CNN) ignited a renewed interest in this field, due to the very impressive results obtained. There exists an alternative path toward handling the style transfer task, via the generalization of texture synthesis algorithms. This approach has been proposed over the years, but its results are typically less impressive compared with the CNN ones. In this paper, we propose a novel style transfer algorithm that extends the texture synthesis work of Kwatra et al. (2005), while aiming to get stylized images that are closer in quality to the CNN ones. We modify Kwatra’s algorithm in several key ways in order to achieve the desired transfer, with emphasis on a consistent way for keeping the content intact in selected regions, while producing hallucinated and rich style in others. The results obtained are visually pleasing and diverse, shown to be competitive with the recent CNN style transfer algorithms. The proposed algorithm is fast and flexible, being able to process any pair of content + style images. [ABSTRACT FROM PUBLISHER]