Search

Face hallucination or super-resolution is a practical application of general image super-resolution which has been recently studied by many researchers. The challenge of good face hallucination comes from a variety of poses, illuminations, facial expressions, and other degradations. In many proposed methods, researchers resolve it by using a generative neural network to reduce the perceptual loss so we can generate a photo-realistic image. The problem is that researchers usually overlook the fidelity of the super-resolved image which could affect further facial image processing. Meanwhile, many CNN based approaches cascade multiple networks to extract facial prior information to improve super-resolution quality. Because of the end-to-end design, the details are missing for investigation. In this paper, we combine new techniques in convolutional neural network and random forests to a Hierarchical CNN based Random Forests (HCRF) approach for face super-resolution in a coarse-to-fine manner. In the proposed approach, we focus on a general approach that can handle facial images with various conditions without pre-processing. To the best of our knowledge, this is the first paper that combines the advantages of deep learning with random forests for face super-resolution. To achieve superior performance, we propose two novel CNN models for coarse facial image super-resolution and segmentation and then apply new random forests to target on local facial features refinement making use of the segmentation results. Extensive benchmark experiments on subjective and objective evaluation show that HCRF can achieve comparable speed and competitive performance compared with state-of-the-art super-resolution approaches for very low-resolution images. [ABSTRACT FROM AUTHOR]

This paper presents a modular push–pull PWM converter (MPC) for a battery energy storage system, which is intended for grid connections to medium- or high-voltage power systems. The converter per phase consists of a center-tapped transformer and two arms based on a cascade connection of multiple bidirectional PWM chopper-cells with floating dc capacitors. This paper discusses the operating performance and control method of the MPC, focusing on voltage balancing of all the floating dc capacitors. Moreover, a comparison is made between the MPC and a modular multilevel converter (MMC), under an assumption that both are used as the same battery energy storage system. The validity of the operating performance and control method is confirmed by both computer simulation using the “PSCAD/EMTDC” software package and experiment using a three-phase, 200-V, 5-kW downscaled model. [ABSTRACT FROM PUBLISHER]

In this paper, real outage data from utilities are analyzed to gain better understanding of cascading outage propagation. In order to accurately estimate the interactions between component outages, two mechanisms are introduced: the evolution of interactions over generations and the memory between consecutive generations. A metric, the expected number of outages following one component outage, is calculated based on the estimated interaction networks by solving a set of carefully formulated linear equations, considering loops due to the complex component interactions. Existence of unique positive solution for the linear equations is mathematically proved. Components that are critical for outage propagation are further identified based on the developed metric. Besides, the outage propagation properties are revealed by the interaction networks estimated from real outage data. Further, the cascades simulated from a highly probabilistic generation-dependent interaction model using the estimated interactions well capture the properties of the original outage data in terms of the distribution of the number of line outages, the offspring mean of the branching process, the distribution of the component metrics, and the identified critical components. Cascading failure risks are greatly mitigated by reducing the probability that the identified critical components fail. [ABSTRACT FROM AUTHOR]

The modular multilevel converter (MMC), owing to the presence of a large number of submodules (SMs) and energy storing elements, requires a large footprint and initial investment. To address these issues, this paper proposes the switched modular multilevel converter (SMMC). It consists of two major elements; the chain-links (CLs), consisting of cascade connection of SMs as in MMC, and the single-pole-triple-throw (SPTT) switches, which can be realized using either the T-type or NPC-type switch strings. The SMMC works on the principle of connecting all the phase-legs in series and their respective positions are mutually interchanged. It offers a compact structure as the SMs count is reduced to 33% of MMC, which results in a much lower energy storage requirement. Furthermore, it does not have inter-phase circulating current problem and hence the CL inductors can also be avoided. The SMMC structure, operating principles, control, and modulation schemes are explained in detail. A detailed comparison of SMMC with the other major existing converters is carried out to highlight the important features of SMMC. The efficacy of SMMC is validated using PSCAD simulation studies with a 19-level grid-connected SMMC model. A 3-phase 19-level SMMC laboratory prototype is developed to further validate its feasibility. [ABSTRACT FROM AUTHOR]

Cascading failure on complex networks has been extensively studied over the past decade. However, restoration of networks from cascading failure is still relatively unexplored. In this paper, we consider cascading failure in conjunction with the restoration process involving repairing the failed nodes in a sequential fashion. Depending on the availability of resources, we tackle the sequential recovery problem from two distinct approaches, namely, result-oriented and resource-oriented restoration approaches. In the result-oriented approach, we aim to restore the network to the largest extent and within the shortest time. Heuristic network restoration strategies based on node load or degree are proposed. For resource-oriented restoration, we aim to maximize the increase of network size with a given number of nodes to be repaired, and we propose a novel iterative strategy to improve performance. Simulation results on the Barábasi–Albert scale-free network, Internet autonomous system-level network, and IEEE 300 bus power system have demonstrated the effectiveness of the proposed sequential recovery strategies. [ABSTRACT FROM AUTHOR]

Due to the low-frequency oscillation (LFO) between traction network and electric multiple units (EMUs) of high-speed railway, many traction blockades occur in different EMU depots. Aiming at this problem, the mathematical model and small-signal analysis model of EMUs-traction network cascade system (ETNCS) are built in this paper. The system stability mainly affected by the electrical coupling between source and load module is analyzed and discussed in detail. Considering the autotransformer leakage reactance, traction network impedance, and distributed capacitance, the output source module resistance of ETNCS is calculated based on the generalized symmetrical component method. Using the generalized Nyquist criterion and Bode diagram, the ETNCS's stability is analyzed and the quantitative indicators of system impedance matching are given. Then, the simulation model of full ETNCS is built in this paper, which can reproduce the LFO phenomenon. In addition, the theoretical analysis results and the feasibility of inhibition for LFO are verified based on MATLAB/simulink simulation. Finally, the LFO signal modal analysis is implemented by using Prony and estimating signal parameters via rotational invariance techniques (ESPRIT) methods. The conclusions show that the proposed stability and modal analysis methods could be used not only to analyze the LFO more clearly, but also to provide the technical reference for the stable operation of EMUs. [ABSTRACT FROM PUBLISHER]

Chronological age of healthy people is able to be predicted accurately using deep neural networks from neuroimaging data, and the predicted brain age could serve as a biomarker for detecting aging-related diseases. In this paper, a novel 3D convolutional network, called two-stage-age-network (TSAN), is proposed to estimate brain age from T1-weighted MRI data. Compared with existing methods, TSAN has the following improvements. First, TSAN uses a two-stage cascade network architecture, where the first-stage network estimates a rough brain age, then the second-stage network estimates the brain age more accurately from the discretized brain age by the first-stage network. Second, to our knowledge, TSAN is the first work to apply novel ranking losses in brain age estimation, together with the traditional mean square error (MSE) loss. Third, densely connected paths are used to combine feature maps with different scales. The experiments with 6586 MRIs showed that TSAN could provide accurate brain age estimation, yielding mean absolute error (MAE) of 2.428 and Pearson’s correlation coefficient (PCC) of 0.985, between the estimated and chronological ages. Furthermore, using the brain age gap between brain age and chronological age as a biomarker, Alzheimer’s disease (AD) and Mild Cognitive Impairment (MCI) can be distinguished from healthy control (HC) subjects by support vector machine (SVM). Classification AUC in AD/HC and MCI/HC was 0.904 and 0.823, respectively. It showed that brain age gap is an effective biomarker associated with risk of dementia, and has potential for early-stage dementia risk screening. The codes and trained models have been released on GitHub: https://github.com/Milan-BUAA/TSAN-brain-age-estimation. [ABSTRACT FROM AUTHOR]

Deep convolutional neural networks (DCNNs) are one of the most promising deep learning techniques and have been recognized as the dominant approach for almost all recognition and detection tasks. The computation of DCNNs is memory intensive due to large feature maps and neuron connections, and the performance highly depends on the capability of hardware resources. With the recent trend of wearable devices and Internet of Things, it becomes desirable to integrate the DCNNs onto embedded and portable devices that require low power and energy consumptions and small hardware footprints. Recently stochastic computing (SC)-DCNN demonstrated that SC as a low-cost substitute to binary-based computing radically simplifies the hardware implementation of arithmetic units and has the potential to satisfy the stringent power requirements in embedded devices. In SC, many arithmetic operations that are resource-consuming in binary designs can be implemented with very simple hardware logic, alleviating the extensive computational complexity. It offers a colossal design space for integration and optimization due to its reduced area and soft error resiliency. In this paper, we present HEIF, a highly efficient SC-based inference framework of the large-scale DCNNs, with broad applications including (but not limited to) LeNet-5 and AlexNet, that achieves high energy efficiency and low area/hardware cost. Compared to SC-DCNN, HEIF features: 1) the first (to the best of our knowledge) SC-based rectified linear unit activation function to catch up with the recent advances in software models and mitigate degradation in application-level accuracy; 2) the redesigned approximate parallel counter and optimized stochastic multiplication using transmission gates and inverse mirror adders; and 3) the new optimization of weight storage using clustering. Most importantly, to achieve maximum energy efficiency while maintaining acceptable accuracy, HEIF considers holistic optimizations on cascade connection of function blocks in DCNN, pipelining technique, and bit-stream length reduction. Experimental results show that in large-scale applications HEIF outperforms previous SC-DCNN by the throughput of $4.1\times $ , by area efficiency of up to $6.5\times $ , and achieves up to ${5.6\times }$ energy improvement. [ABSTRACT FROM AUTHOR]

Recently, deep learning techniques have substantially boosted the performance of salient object detection in still images. However, the salient object detection in videos by using traditional handcrafted features or deep learning features is not fully investigated, probably due to the lack of sufficient manually labeled video data for saliency modeling, especially for the data-driven deep learning. This paper proposes a novel weakly supervised approach to the salient object detection in a video, which can learn a robust saliency prediction model by using very limited manually labeled data and a large amount of weakly labeled data that could be easily generated in a supervised approach. Furthermore, we propose a spatiotemporal cascade neural network architecture for saliency modeling, in which two fully convolutional networks are cascaded to evaluate the visual saliency from both spatial and temporal cues to lead the optimal video saliency prediction. The proposed approach is extensively evaluated on the widely used challenging data sets, and the experiments demonstrate that our proposed approach substantially outperforms the state-of-the-art salient object detection models. [ABSTRACT FROM AUTHOR]

In this paper, a newly shunted-line stepped-impedance resonator (SLSIR) is studied and applied to realize high-order balanced filter design for radio astronomy application. The proposed SLSIR not only has more flexible control of the internal coupling between adjacent units but also provides more design freedom for the cascade connection of high-order filter. Then, the operating theory of three kinds of electrically, magnetically, and electromagnetically coupled SLSIR pairs has been discussed and utilized to high-performance filter design. Hence, fourth-order balanced filter using electromagnetically coupled SLSIR pairs is investigated and designed, which exhibits an excellent filtering performance under differential-mode (DM) operation as well as a high common-mode (CM) suppression level. In addition, the production mechanism of DM transmission zero is modeled and analyzed. Based on the analysis above, two dissimilar types of coupled SLSIR pairs with the same fundamental frequency but different harmonics are adopted to design an eighth-order balanced filter. Finally, the high-order balanced filter is fabricated on 0.5-mm-thick MgO wafer with double-sided YBCO thin films. Measured results show that the eighth-order balanced filter operates at 2.53 GHz with the corresponding in-band insertion loss of 0.1 dB. Results well verify the proposed structure with high selectivity, low insertion loss, and wide stopband performances. [ABSTRACT FROM AUTHOR]

When two filters are cascaded together, unwanted spikes appear in the rejection region. These spikes can have a strong impact on the rejection performance of the resulting filter. Therefore, it is desired to suppress or move them far away from the operational band, including part of the rejection band as well. This paper addresses a systematic evaluation of such phenomenon and proposes a simple yet effective method, called spike envelope curve, to have, in a single run and for any frequency, the worst possible case, without the need of repeating onerous analyses. It is also shown that a judicious choice of the length and losses of the line separating the filters can be exploited to mitigate considerably such a negative effect. The method is assessed by full-wave simulations and measurements. [ABSTRACT FROM AUTHOR]

Face recognition (FR) is a widely studied topic in the field of computer vision research. Although promising results are achieved, FR researches still face challenges of age variations. Most existing FR networks conduct classification based on the feature similarity, which may be misled by large intra class difference under age variations. Instead of calculating feature similarity, in this paper, we derive a novel cross-age face verification framework named Cross-Age Identity Difference Analysis (CIDA) model, which analyzes the identity difference between image pairs under age variations. Specifically, our framework includes two cascading networks. Firstly, an Identity Difference Feature Extractor (IDFE) is proposed to extract the difference information between the input image pair, where the identity discriminant features are effectively extracted, while other interference factors such as age, illumination, posture are suppressed. Secondly, the Direct Cross-age Verification Network (DCVN) is proposed to directly decide whether the input image pair is from the same individual. We derive a novel loss function, where the classification loss with larger age difference is assigned larger weights, which urges the classifier to pay attention to the classification process of the samples with large age gap. Besides, the loss of DCVN are integrated with the loss function of IDFE as a feedback of the final classification performance, improving the discriminant power of IDFE. Through synchronous training of the two networks, we can finally achieve end-to-end network architecture. Compared with the existing cross-age face recognition (CAFR) methods, we do not need to consider feature similarity comparison, which provides a new insight for cross-age face recognition task. Extensive experiments have been performed on the benchmark CAFR datasets which verify the effectiveness of our model. [ABSTRACT FROM AUTHOR]

Nonlinear feedback shift registers (NFSRs) are widely used in stream cipher design as building blocks. In this paper, we study the problem of decomposing an NFSR into the cascade connection of an NFSR into a linear feedback shift register (LFSR), which is a kind of concatenation of an NFSR and LFSR. A necessary and sufficient condition for such decomposition is provided and other algebraic properties about such decomposition are also studied. Based on these theoretical results, a binary decision diagram (BDD)-based algorithm for such decomposition is proposed. Compared with the previous algorithm proposed by Ma et al., our algorithm can find more accurate candidate LFSR and the algebraic properties presented in this paper guarantee that the memory requirement during our verification is linear in the size of the BDD of the NFSRs characteristic function. [ABSTRACT FROM PUBLISHER]

This paper proposes a circular polarization selective surface (CPSS) design consisting of two meander-line circular polarizers and a linear polarizer. No vias are required. These polarizers are individually designed with ANSYS-HFSS (High Frequency Structure Simulator) software whereas the response of the overall cascade CPSS is computed with the GSM approach. A cascade CPSS which includes three-layer meander-line polarizers is designed and fabricated. Good agreement is shown between experimental and simulated results, with only the TE00 and TM00 modes in the GSM analysis. Finally, analysis of the circular polarizers based on simple filter theory is done to better understand how the performance of the whole structure can be improved. [ABSTRACT FROM PUBLISHER]

A Grain-like structure is a cascade connection of a primitive LFSR into an NFSR. It is well known that such structures generate sequences with periods multiples of the period of primitive LFSR sequences. In this paper, we study weak Grain-like structures, i.e., Grain-like structures generating at least one sequence with minimum period. Assume the orders of LFSR and NFSR are $n$ and $m$ respectively. We prove that weak Grain-like structures always exist when $m>n$. For $m=n$ , we give three classes of weak Grain-like structures. Then we extend the method in the second class to prove that weak Grain-like structures exist when $m\ge n-\lg n+4$. Moreover, our experimental data shows that, the ratio of weak Grain-like structures approximates a value which is a little more than 63% for small $m=n$ , and weak Grain-like structures exist with $m=3$ or 4 when $n\le 18$. [ABSTRACT FROM AUTHOR]

Image captioning, a challenging task where the machine automatically describes an image with natural language, has drawn significant attention in recent years. Despite the remarkable improvements of recent approaches, however, these methods are built upon a large set of training image-sentence pairs. The expensive labor efforts hence limit the captioning model to describe the wider world. In this paper, we present a novel network structure, Cascaded Revision Network, which aims at relieving the problem by equipping the model with out-of-domain knowledge. CRN first tries its best to describe an image using the existing vocabulary from in-domain knowledge. Due to the lack of out-of-domain knowledge, the caption may be inaccurate or include ambiguous words for the image with unknown (novel) objects. We propose to re-edit the primary captioning sentence by a series of cascaded operations. We introduce a perplexity predictor to find out which words are most likely to be inaccurate given the input image. Thereafter, we utilize external knowledge from a pretrained object detection model and select more accurate words from detection results by the visual matching module. In the last step, we design a semantic matching module to ensure that the novel object is fit in the right position. By this novel cascaded captioning-revising mechanism, CRN can accurately describe images with unseen objects. We validate the proposed method with state-of-the-art performance on the held-out MSCOCO dataset as well as scale to ImageNet, demonstrating the effectiveness of our method. [ABSTRACT FROM AUTHOR]

In this paper we address the problem of hallucinating high-resolution facial images from low-resolution inputs at high magnification factors. We approach this task with convolutional neural networks (CNNs) and propose a novel (deep) face hallucination model that incorporates identity priors into the learning procedure. The model consists of two main parts: i) a cascaded super-resolution network that upscales the low-resolution facial images, and ii) an ensemble of face recognition models that act as identity priors for the super-resolution network during training. Different from most competing super-resolution techniques that rely on a single model for upscaling (even with large magnification factors), our network uses a cascade of multiple SR models that progressively upscale the low-resolution images using steps of $2\times $. This characteristic allows us to apply supervision signals (target appearances) at different resolutions and incorporate identity constraints at multiple-scales. The proposed C-SRIP model (Cascaded Super Resolution with Identity Priors) is able to upscale (tiny) low-resolution images captured in unconstrained conditions and produce visually convincing results for diverse low-resolution inputs. We rigorously evaluate the proposed model on the Labeled Faces in the Wild (LFW), Helen and CelebA datasets and report superior performance compared to the existing state-of-the-art. [ABSTRACT FROM AUTHOR]

This paper provides a theoretical and experimental discussion about the performance, availability, and flexibility of a battery energy storage system (BESS) using a modular multilevel cascaded converter based on single-star bridge cells (MMCC-SSBC). The SSBC-based BESS produces three-phase multilevel voltage waveforms, and eliminates both harmonic filters and a complicated zig-zag transformer from the ac side. The circuit modularity of the SSBC enables the usage of multiple individual low-voltage battery modules. Along with control strategy based on zero-sequence-voltage injection, this modularity enhances the availability and flexibility of the BESS. A three-phase laboratory downscaled system rated at 140 V, 10 kW, and 21 kW·h is designed, constructed, and tested to verify the operating principles and performance. The tested BESS produces a current total harmonic distortion (THD) of 3.0% even when it operates with different active-power commands for individual bridge cells. In addition, zero-voltage ride-through capability is verified for the severest single-phase, two-phase, and three-phase voltage sags. Mathematical analysis and experimental verification validate the downscaled system and performance, making the BESS prospective. [ABSTRACT FROM PUBLISHER]

A set of cascaded inductive voltage dividers (IVDs) with a ratio of $2^{n}$ :1 and resistive voltage dividers (RVDs) with serial–parallel connection structure is presented in this paper. The IVD with the ratio of $2^{n}$ :1 is built with a $2^{n-1}$ :1 IVD and a binary IVD connected in the cascade structure. The phase angle errors of the IVDs are calibrated in a step-up procedure by using the three BIVDs in cascade connection as the reference standard. The phase angle errors of the RVDs can also be self-calibrated. The self-calibration results and uncertainties of phase angle errors of the IVDs and RVDs have been both given at frequencies up to 100 kHz. A closure measurement between the IVD with the ratio of 256:1 and the RVD with the ratio of 257:1 has also been done at frequencies from 25 to 100 kHz. The closure results show good agreement within 15 $\mu $ rad compared to the self-calibration results. [ABSTRACT FROM AUTHOR]

Internet of Vehicles (IoV) has become a significant research area due to its specific features and applications such as efficient traffic management, road safety, and entertainment. Vehicles are expected to carry relatively more communication systems, onboard computing facilities, storage, and increased sensing power. Vehicular Cloud Network (VCN) is a hybrid technology that has a remarkable impact on IoV by instantly using vehicular resources. Merging software-defined networking with VCN is valuable to vehicular networks. Network function virtualization can be deployed across the servers and vehicles within the VCN to provide customized agile services. Its deployment requires an appropriate management strategy that often manifests as the difficult online decision making tasks. This paper proposes an intelligent Virtualized Network Functions (VNFs) selection strategy. For satisfying the quality requirements of different vehicular services, this strategy is based on the results of traffic identification that utilizes deep neural network and Multi-Grained Cascade Forest to distinguish service behaviors. According to the order of VNFs, the vehicular network is divided into several layers where each arrived packet needs to be queued. The scheduler of each layer selects a layer to host the next VNF for the packets in the queue. Our experiments demonstrate that the proposed traffic identification method increases the precision by 7.7% and improves the real-time performance. Furthermore, the proposed model of VNFs selection reduces network congestion compared to traditional scheduling models. [ABSTRACT FROM AUTHOR]

Goal: Automatic segmentation of brainstem gliomas and prediction of genotype (H3 K27M) mutation status based on magnetic resonance (MR) images are crucial but challenging tasks for computer-aided diagnosis in neurosurgery. In this paper, we present a novel cascaded deep convolutional neural network (CNN) to address these two challenging tasks simultaneously. Methods: Our novel segmentation task contains two feature-fusion modules: the Gaussian-pyramid multiscale input features-fusion technique and the brainstem-region feature enhancement. The aim is to resolve very difficult problems in brainstem glioma segmentation. Our prediction model combines CNN features and support-vector-machine classifier to automatically predict genotypes without region-of-interest labeled-MR images and is learned jointly with the segmentation task. First, Gaussian-pyramid multiscale input feature fusion is added to our glioma-segmentation task to solve the problems of size variety and weak brainstem-gliomas boundaries. Second, the two feature-fusion modules provide both local and global contexts to retain higher frequency details for sharper tumor boundaries, handling the problem of the large variation of tumor shape, and volume resolution. Results and Conclusion: Experiments demonstrate that our cascaded CNN method achieves not only a good tumor segmentation result with a high Dice similarity coefficient of 77.03%, but also a competitive genotype prediction result with an average accuracy of 94.85% upon fivefold cross-validation. [ABSTRACT FROM AUTHOR]

Considering cache enabled networks, optimal content placement minimizing the total cost of communication in such networks is studied, leading to a surprising fundamental 0–1 law for non-redundant cache placement strategies, where the total cache sizes associated with each file does not exceed the file size. In other words, for such strategies, we prove that any non-redundant cache placement strategy can be transformed, with no additional cost, to a strategy in which at every node, each file is either cached completely or not cached at all. Moreover, we obtain a sufficient condition under which the optimal cache placement strategy is in fact non-redundant. This result together with the 0–1 law reveals that situations exist, where optimal content placement is achieved just by uncoded placement of whole files in caches. [ABSTRACT FROM PUBLISHER]

Aiming at the problems of complex background, diverse shapes, and object occlusion in aerial images, a cascade reasoning graph network (CRGN) is proposed for multi-fitting detection on transmission lines. First of all, for these three problems mentioned above, co-occurrence knowledge, semantic knowledge, and spatial knowledge were constructed to represent the co-relation of objects by analyzing the characteristics of the transmission line fittings. Next, the Supervised Graph Learning (SGL), Graph Attention network (GAT), and Graph Convolutional Network (GCN) were employed to reason corresponding knowledge. In addition, to generate more accurate proposals for the graph reasoning module, resampling was carried out through the cascade network. Finally, the enhanced features were fused with the original visual features to recognize and position the fittings. Test results show that CRGN can improve the detection effect of multi-fittings on the transmission line, especially for some hard-detection fittings. [ABSTRACT FROM AUTHOR]

This article presents a ground breaking traction drive for fuel cell-powered light rail vehicles based on a multilevel cascade converter with H-bridge cells. The converter provides dc–ac power conversion with unique buck–boost capability to compensate for the variation of fuel cell terminal voltage with the load power. In comparison to a conventional boost inverter, based on the cascade connection of a boost dc–dc converter and a voltage source inverter, the proposed converter has the significant benefits of much smaller inductor size, lower dc side current ripple, and multilevel voltage waveforms at the ac output. This article presents the converter modeling and design of the boost inductor and submodule capacitors, taking into account the full speed range of the motor. The concept has been fully validated on a bespoke experimental prototype driving an induction motor, showing the suitability of the proposed converter as a traction drive for hydrogen-powered rail vehicles. [ABSTRACT FROM AUTHOR]

Video referring segmentation focuses on segmenting out the object in a video based on the corresponding textual description. Previous works have primarily tackled this task by devising two crucial parts, an intra-modal module for context modeling and an inter-modal module for heterogeneous alignment. However, there are two essential drawbacks of this approach: (1) it lacks joint learning of context modeling and heterogeneous alignment, leading to insufficient interactions among input elements; (2) both modules require task-specific expert knowledge to design, which severely limits the flexibility and generality of prior methods. To address these problems, we here propose a novel Object-Agnostic Transformer-based Network, called OATNet, that simultaneously conducts intra-modal and inter-modal learning for video referring segmentation, without the aid of object detection or category-specific pixel labeling. More specifically, we first directly feed the sequence of textual tokens and visual tokens (pixels rather than detected object bounding boxes) into a multi-modal encoder, where context and alignment are simultaneously and effectively explored. We then design a novel cascade segmentation network to decouple our task into coarse-grained segmentation and fine-grained refinement. Moreover, considering the difficulty of samples, a more balanced metric is provided to better diagnose the performance of the proposed method. Extensive experiments on two popular datasets, A2D Sentences and J-HMDB Sentences, demonstrate that our proposed approach noticeably outperforms state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In recent years, marine animal study has gained increasing research attention, which raises significant demands for fine-grained marine animal segmentation (MAS) techniques. In addition, deep learning has been widely adopted for object segmentation and has achieved promising performance. However, deep-based MAS is still lack of investigation due to the shortage of a large-scale MAS dataset. To tackle this issue, we construct the first large-scale MAS dataset, called MAS3K, which consists of 3,103 images from different types, including camouflaged marine animal images, common marine animal images, and underwater images without marine animals. Furthermore, we consider different underwater conditions, such as low illumination, turbid water quality, photographic distortion, etc. Each image from MAS3K dataset has rich annotations, including an object-level mask, a category name, attributes, and a camouflage method (if applicable). Furthermore, we propose a novel MAS network, called Enhanced Cascade Decoder Network (ECD-Net), which consists of multiple Interactive Feature Enhancement Modules (IFEMs) and Cascade Decoder Modules (CDMs). In ECD-Net, the IFEMs are first utilized to extract rich multi-scale features. The resulting features are then fed to the CDMs for accurately segmenting marine animals from complex underwater environments. We perform extensive experiments to compare ECD-Net with 10 cutting-edge object segmentation models. The results demonstrate that ECD-Net is an effective MAS model and outperforms the cutting-edge models, both qualitatively and quantitatively. [ABSTRACT FROM AUTHOR]

Super-Resolution convolutional neural networks have recently demonstrated high-quality restoration for single images. However, existing algorithms often require very deep architectures and long training times. Furthermore, current convolutional neural networks for super-resolution are unable to exploit features at multiple scales and weigh them equally or at only static scale only, limiting their learning capability. In this exposition, we present a compact and accurate super-resolution algorithm, namely, densely residual laplacian network (DRLN). The proposed network employs cascading residual on the residual structure to allow the flow of low-frequency information to focus on learning high and mid-level features. In addition, deep supervision is achieved via the densely concatenated residual blocks settings, which also helps in learning from high-level complex features. Moreover, we propose Laplacian attention to model the crucial features to learn the inter and intra-level dependencies between the feature maps. Furthermore, comprehensive quantitative and qualitative evaluations on low-resolution, noisy low-resolution, and real historical image benchmark datasets illustrate that our DRLN algorithm performs favorably against the state-of-the-art methods visually and accurately. [ABSTRACT FROM AUTHOR]

Network noise invariants are introduced that lead to improved noise characterization and a complete theory of linear noisy two-ports. Minimum power-added noise temperature and minimum cold load temperature are identified as network noise invariants under lossless embedding. Associated invariant equations provide explicit relations between all known and new network invariants. From these equations, an invariant under lossless embedding is identified that defines network noise–gain coupling in the most basic terms of noise correlation, minimum noise temperature, and complex nonreciprocal gain. A noise correlation parameter q is formally introduced that is invariant to lossless input and/or output transformation. Conditions and bounds are established, and it is shown that q \approx 2 for low-noise active devices. An exact expression for the q parameter of a minimum noise cascade network is given in terms of constituent device invariants. From a systems point of view, the cascade q parameter represents source impedance noise sensitivity. A lower bound on cascade q is determined by device invariants minimum power-added noise temperature and minimum cold load temperature. It is shown that the cascade q lower bound is realized by simultaneous noise and power match. [ABSTRACT FROM PUBLISHER]

In recent years, deep learning (DL) has attracted increasing attention in hyperspectral unmixing (HU) applications due to its powerful learning and data fitting ability. The autoencoder (AE) framework, as an unmixing baseline network, achieves good performance in HU by automatically learning low-dimensional embeddings and reconstructing data. Nevertheless, the conventional AE-based architecture, which focuses more on the pixel-level reconstruction loss, tends to lose some significant detailed information of certain materials (e.g., material-related properties) in the reconstruction process. Therefore, inspired by the perception mechanism, we propose a cycle-consistency unmixing network, called CyCU-Net, by learning two cascaded AEs in an end-to-end fashion, to enhance the unmixing performance more effectively. CyCU-Net is capable of reducing the detailed and material-related information loss in the process of reconstruction by relaxing the original pixel-level reconstruction assumption to cycle consistency dominated by the cascaded AEs. More specifically, cycle consistency can be achieved by a newly proposed self-perception loss, which consists of two spectral reconstruction terms and one abundance reconstruction term. By taking advantage of the self-perception loss in the network, the high-level semantic information can be well preserved in the unmixing process. Moreover, we investigate the performance gain of CyCU-Net with extensive ablation studies. Experimental results on one synthetic and three real hyperspectral data sets demonstrate the effectiveness and competitiveness of the proposed CyCU-Net in comparison with several state-of-the-art unmixing algorithms. [ABSTRACT FROM AUTHOR]

Crowd management is an essential task to ensure the safety and smoothness of any event. Using novel technologies, including surveillance cameras, drones, and the communication techniques between security agents, the control of the crowd has become easier. Yet, the usage of such techniques is still not effective. This article presents an approach for crowd counting from drones' data. The proposed method exploits the dilated and scaled neural networks for feature extraction and density crowd estimation. A new dataset named ViseDrone2020 is used for training and testing of the proposed method. In order to compare the proposed method, we implemented 10 state-of-the-art methods and trained it on this dataset. The experiments show that the proposed model is more efficient for crowd counting compared with the implemented methods. Also, some of these methods give relatively accurate results in terms of estimated crowd numbers as well as the quality of estimated density maps. The proposed model was further evaluated on nondrone datasets, namely, UCF_QNRF, UCF_CC_50, and shanghaiTech_(A, B), which produced satisfying results for all the datasets. In addition, the proposed method was tested on noisy images where Gaussian noise and salt-and-pepper noise was applied to all the images of the dataset with a noise density of 0.02. The analysis showed that the quality of the density map as well as the quantity of the crowd count estimation is comparatively better than other existing methods without the presence of noise. After the acceptance, the code of the proposed method as well as 10 implemented methods will be available at: [Online]. Available: https://github.com/elharroussomar/Crowd-Conting-on-Drone-Data. [ABSTRACT FROM AUTHOR]

This article presents 16-times frequency multiplier composing of two four-times frequency multipliers (quadrupler) in cascade connection. The proposed topology is new structure of tuned-frequency multiplier (TFM) for better harmonic rejection ratio (HRR) with wide frequency range and low power consumption. For accomplishing reliable output frequency band of 16–28 GHz, the whole band is divided into 64 subsidiary frequency bands by applying 6-bits digitally controlled capacitor-bank of LC-tuned tank. The proposed quadrupler is consisted of a harmonic generator (HG) and a cascode LC-tuned buffer. The proposed HG topology is based on double balanced mixer (DBM). Unlike typical mixer bias, the bottom differential pair devices of the proposed HG are C-class biased to generate more desired the fourth order harmonic. In addition, to reduce power consumption and frequency conversion gain variations for the whole target frequency band, negative- $g_{m}$ differential pair is added in parallel to enhance the equivalent parasitic parallel resistance of the HG LC-tank while keep it from oscillation. Great efforts have been contributed to minimize process variation effects by simple but relatively accurate capacitance calibration. Furthermore, each LC-tuned tank output amplitude is regulated by a loop to maintain same output swing for the best optimization of specifications such as power consumption and HRR. The proposed 16-times multiplier is fabricated on 65-nm complementary metal–oxide–semiconductor (CMOS) process and successfully tested. Chip die size 0.7 mm2 excluding input/output (I/O) pads and average power consumption is only 6 mW for 16–28 GHz frequency band. Also, negligible phase noise degradation is achieved. [ABSTRACT FROM AUTHOR]

Cross Age Face Recognition (CAFR) is a challenging task in the field of face recognition. There still exist some limitations in mainstream CAFR methods. On one hand, some methods need synthesizing multiple groups of features and fusing the results under different ages, where the similarity score in some age groups may affect final classification results. On the other hand, many other methods takes a facial image as a linear combination of identity information and age information, which treats age factor as a value independent of identity information, but may be inconsistent with the aging pattern of many individuals. And these methods require both age labels and identity labels in training, which is limited by the scale of existing CAFR datasets. To address the above limitations, this work proposes the Parallel Multi-path Age Distinguish Network (PMADN) model. Specifically, our model consists of two cascading networks, an Age Distinguish Mapping Network (ADMN) and a Cross-Age Feature Recombination Network (CFRN). Firstly, the face features are mapped into different age groups by parallel multi-path full connected layers in ADMN, which can better extract the identity features in a small age span. Secondly, CFRN nonlinearly recombines the mapped features to extract the age robust features that are beneficial to identity classification, which can avoid the simple linear combination of identity factor and age factor in the existing methods. What’s more, our algorithm is combined with transfer learning and only uses the age label and a pre-trained ordinary face recognition network for training, which can make use of a larger aging face dataset for training. Extensive CAFR experiments performed on the benchmark MORPH Album2, CACD-VS and Cross Age LFW databases demonstrate the effectiveness and superiority of our method. [ABSTRACT FROM AUTHOR]

Intrinsic image analysis plays an important role for image understanding, since it can provide accurate reflectance, shape and illumination information of the scene. However, intrinsic image analysis is an ill-posed problem which need to apply extra constrains for the decomposition of reflectance image and shading image from a single image. Recently deep neural networks are introduced for intrinsic image analysis, which can produce two intrinsic components simultaneously. In fact, the mutually exclusive relationship between reflectance image and shading image is not only a constraint for decomposition but also can improve the decomposition results. However, this relationship is always omitted in the current networks. In order to address this problem, we propose a novel deep network called as Cascaded Quotient Network (CasQNet) for intrinsic image decomposition. The CasQNet consists of two sub-networks: a Pyramid Mini-U-Net (PyNet) that specifically extracts the reflectance image in multi-scale and a Shading Optimization Network (SoNet) that optimizes the resulting shading. These two sub-networks are cascaded by a quotient operation, which directly enforces the mutually exclusive relationship between reflectance image and shading image in the network architecture. In PyNet, the task of reconstructing reflectance image is achieved by a series of nested multi-scale U-Nets, which simplified the learning task for each U-Net. SoNet is designed to address the unsmooth and blur problems of extreme points caused by the quotient operation. PyNet and SoNet are trained alternately and finally jointed in cascaded structure. Furthermore, we combine multiple loss functions, which consist of data loss, correlation loss and reconstruction loss, for improving the learning effectiveness. To evaluate our proposed algorithm, extensive experiments are performed on three datasets, i.e., ShapeNet, BOLD Surface and MIT Intrinsic Image datasets. Qualitative and quantitative results show that our model achieves the best performance compared to the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Three-dimensional object detection plays an important role in autonomous driving because it provides the 3-D locations of objects for subsequent use in decision-making modules. A novel method is proposed using a monocular image and cascade geometric constraints to achieve robust 3-D vehicle detection. The framework is divided into two stages. In the first stage, the monocular image input is processed using a heatmap-based detection network with five branches to regress the orientation, dimension, center projection of the bottom face, viewpoint classification, and 2-D bounding box. In the second stage, the intersection-over-union threshold is increased to filter out imprecise 2-D bounding boxes. Thereafter, cascade geometric constraints are used to obtain the final 3-D box output, which improves the detection performance under truncation and occlusion conditions. The proposed method is tested on the KITTI-3-D benchmark and is shown to be effective and efficient. The proposed framework does not depend on external sources or subnetworks and can be trained in an end-to-end manner. [ABSTRACT FROM AUTHOR]

This article presents an interleaved bidirectional chopper with auxiliary converters for onboard energy storage systems that is capable of achieving a significant reduction in the mass and volume of the current-smoothing inductors. The proposed chopper consists of multiple interleaved subconverters, where each subconverter consists of a bidirectional chopper cell, a current-smoothing inductor at the low-voltage side, and an auxiliary converter that is formed by the cascade connection of multiple full-bridge cells. Each auxiliary converter is connected in series with its respective inductor to suppress the ac voltage component that produces ripple current in the inductor of each subconverter; hence, a smaller inductor can be used to meet the ripple current requirement when compared with the conventional bidirectional chopper. In addition, auxiliary converters can be used as dc-circuit breakers to protect the low-voltage side against faults in the power devices of the bidirectional chopper cells. This article discusses the operating principles and control method of the proposed chopper, followed by experiments using a downscaled experimental model. [ABSTRACT FROM AUTHOR]

The study of information diffusion or cascade has attracted much attention over the last decade. Most related works target on studying cascade-level macroscopic properties such as the final size of a cascade. Existing microscopic cascade models which focus on user-level modeling either make strong assumptions on how a user gets infected by a cascade or limit themselves to a specific scenario where “who infected whom” information is explicitly labeled. The strong assumptions oversimplify the complex diffusion mechanism and prevent these models from better fitting real-world cascade data. Also, the methods which focus on specific scenarios cannot be generalized to a general setting where the diffusion graph is unobserved. To overcome the drawbacks of previous works, we propose a Neural Diffusion Model (NDM) for general microscopic cascade study. NDM makes relaxed assumptions and employs deep learning techniques including attention mechanism and convolutional network for cascade modeling. Both advantages enable our model to go beyond the limitations of previous methods, better fit the diffusion data and generalize to unseen cascades. Experimental results on diffusion identification task over four realistic cascade datasets show that our model can achieve a relative improvement up to 26 percent against the best performing baseline in terms of F1 score. [ABSTRACT FROM AUTHOR]

It is well known that the interaction between cascaded individually designed power conversion systems can cause instability. To overcome this issue, a Hamiltonian energy control scheme is proposed, which is based on passivity control theory and port-controlled Hamiltonian framework. A complementary PI adjustment term is also included in the control algorithm to eliminate the steady-state output voltage error caused by the parameter uncertainty. The proposed control approach is applied to three different cascade structures. First, the cascade structure between dc/dc converters is considered, and the detailed controller design is given. Second, the cascade connection of a single converter and its LC filter is studied. By placing the LC filter into the Hamiltonian model of the controlled converter system, the dynamic and potential instability caused by the filter can be adjusted. Finally, the cascade structure between subsystems including filters and converters, which are common in microgrids, is studied. By using the Hamiltonian function (storage function) as the Lyapunov function candidate, the large-signal stability of each controlled converter system is proved. When the cascade structure contains multiple controlled converter systems, the stability of the entire cascaded system is guaranteed by the superposition of multiple Lyapunov functions. A 3.5 kW 220−270−350 V test bench is built in the laboratory to demonstrate the application of the proposed control approach to these three cascade structures. [ABSTRACT FROM AUTHOR]

The series connection of low-power modules is an alternative solution to realize a multilevel inverter. Unlike a cascade connection, the series connection approach significantly reduces the number of isolated dc sources, thereby the design complexity, cabling of the transformer, and overall system cost become low. In this article, the operation of the series-connected multilevel inverter is presented for a five-level operation, and it is realized by using two three-level half-bridge diode clamp converter modules per phase. Hence, each phase of the series-connected multilevel inverter requires a single isolated dc source. In the series-connected multilevel inverter, the net dc-bus voltage will be equally distributed between four dc-bus capacitors. Therefore, the converter generates a multilevel voltage waveform with uniform steps and ensures equal voltage stress on the semiconductor devices. To achieve these objectives, a space vector pulsewidth modulation scheme with an additional voltage balancing approach is employed. The feasibility of series-connected multilevel inverter is validated through simulations and a scaled-down laboratory prototype under steady-state and transient conditions. [ABSTRACT FROM AUTHOR]

In this work, we propose a motion-guided cascaded refinement network for video object segmentation. By assuming the foreground objects show different motion patterns from the background, for each video frame we apply an active contour model on optical flow to coarsely segment the foreground. The proposed Cascaded Refinement Network (CRN) then takes as guidance the coarse segmentation to generate an accurate segmentation in full resolution. In this way, the motion information and the deep CNNs can complement each other well to accurately segment the foreground objects from video frames. To deal with multi-instance cases, we extend our method with a spatial-temporal instance embedding model that further segments the foreground regions into instances and propagates instance labels. We further introduce a single-channel residual attention module in CRN to incorporate the coarse segmentation map as attention, which makes the network effective and efficient in both training and testing. We perform experiments on popular benchmarks and the results show that our method achieves state-of-the-art performance with high time efficiency. [ABSTRACT FROM AUTHOR]

As the videos captured by hand-held cameras are often perturbed by high-frequency jitters, stabilization of these videos is an essential task. Many video stabilization methods have been proposed to stabilize shaky videos. However, most methods estimate one global homography or several homographies based on fixed meshes to warp the shaky frames into their stabilized views. Due to the existence of parallax, such single or a few homographies can not well handle the depth variation. In contrast to these traditional methods, we propose a novel video stabilization network, called PWStableNet, which comes up pixel-wise warping maps, i.e., potentially different warping for different pixels, and stabilizes each pixel to its stabilized view. To our best knowledge, this is the first deep learning based pixel-wise video stabilization. The proposed method is built upon a multi-stage cascade encoder-decoder architecture and learns pixel-wise warping maps from consecutive unstable frames. Inter-stage connections are also introduced to add feature maps of a former stage to the corresponding feature maps at a latter stage, which enables the latter stage to learn the residual from the feature maps of former stages. This cascade architecture can produce more precise warping maps at latter stages. To ensure the correct learning of pixel-wise warping maps, we use a well-designed loss function to guide the training procedure of the proposed PWStableNet. The proposed stabilization method achieves comparable performance with traditional methods, but stronger robustness and much faster processing speed. Moreover, the proposed stabilization method outperforms some typical CNN-based stabilization methods, especially in videos with strong parallax. Codes will be provided at https://github.com/mindazhao/pix-pix-warping-video-stabilization. [ABSTRACT FROM AUTHOR]

To investigate the harmonic resonance in electric railways, determination of the traction network impedance using a utility instrument is an important and challenging task. This article presents a novel industrial test device which can directly measure the impedance of the actual traction network which is the power source of the railway. The structure arrangement and the parameter design are accomplished to meet the requirements for high-voltage level and strong anti-electromagnetic interference capability. Following this, frequency-dependent impedance measurement is implemented through the injection of controlled harmonic currents using the cascaded H-bridge converters. Furthermore, the power flow analysis is carried out to demonstrate that this device does not need any additional power source and can generate harmonic power. In a new-built railway, this device has been adopted to conduct the on-site test of the traction network impedance. Then, the multi-conductor simulation model is established for the purpose of mutual verification between the field test result and simulation result. Furthermore, the on-site test results can be applied to determine the resonance frequency before a nenw-built line is put into use so that any harmonic resonance can be avoided by taking effective measures in advance. [ABSTRACT FROM AUTHOR]

In this article, we introduce a synthesis-based design approach for waveguide filters, including singlets. As is known, a singlet block comprises one resonator coupled to source and load plus an additional coupling between source and load. In this way, a pole–zero pair is produced, thus allowing the introduction of a transmission zero in the frequency response. Structure making that uses cascaded singlets (especially implemented in waveguide technology) has been successfully used to realize compact and quasi-in-line pseudoelliptic filters. The design of this configuration suffers, however, from a severe drawback; while an isolated singlet can be exactly synthesized, when these blocks are put in cascade, nonresonating nodes (NRNs) are required to allow the cascade connection. Consequently, the resulting topology includes both cross couplings and NRNs, for which no exact synthesis solution is till now available (all the design solutions proposed in the literature are based on optimization). The new design procedure here proposed overcomes this drawback, allowing the synthesis of the equivalent circuit of the cascaded-singlet topology. Once this circuit is obtained, the dimensioning of the singlets can be carried out by exploiting a full-wave simulator according to the procedure described in the article. The proposed design approach allows a very accurate initial dimensioning of the filter, which represents an excellent starting point for final adjustments/optimizations. A detailed design example is presented in the article showing how to practically implement the novel procedure. This new approach is also experimentally validated by a waveguide filter prototype, previously designed by optimization and fabricated. It is shown how this filter can be redesigned with the novel procedure, obtaining very similar physical dimensions in a noticeably shorter time. [ABSTRACT FROM AUTHOR]

As a widely used imaging modality in the medical field, ultrasound has been applied in community medicine, rural medicine, and even telemedicine in recent years. Therefore, the development of portable ultrasound devices has become a popular research topic. However, the limited size of portable ultrasound devices usually degrades the imaging quality, which reduces the diagnostic reliability. To overcome hardware limitations and improve the image quality of portable ultrasound devices, we propose a novel generative adversarial network (GAN) model to achieve mapping between low-quality ultrasound images and corresponding high-quality images. In contrast to the traditional GAN method, our two-stage GAN that cascades a U-Net network prior to the generator as a front end is built to reconstruct the tissue structure, details, and speckle of the reconstructed image. In the training process, an ultrasound plane-wave imaging (PWI) data-based transfer learning method is introduced to facilitate convergence and to eliminate the influence of deformation caused by respiratory activities during data pair acquisition. A gradual tuning strategy is adopted to obtain better results by the PWI transfer learning process. In addition, a comprehensive loss function is presented to combine texture, structure, and perceptual features. Experiments are conducted using simulated, phantom, and clinical data. Our proposed method is compared to four other algorithms, including traditional gray-level-based methods and learning-based methods. The results confirm that the proposed approach obtains the optimum solution for improving quality and offering useful diagnostic information for portable ultrasound images. This technology is of great significance for providing universal medical care. [ABSTRACT FROM AUTHOR]

Two separate transmitarrays that operate at 77 GHz are designed and fabricated. The first transmitarray acts as a quarter-wave plate that transforms a linearly polarized incident wave into a circularly polarized transmitted wave. The second transmitarray acts as both a quarter-wave plate and a beam refracting surface to provide polarization and wavefront control. When the second transmittarray is illuminated with a normally incident, linearly polarized beam, the transmitted field is efficiently refracted to 45^\circ, and the polarization is converted to circular. The half-power bandwidth was measured to be 17%, and the axial ratio of the transmitted field remained below 2.5 dB over the entire bandwidth. Both designs have a subwavelength thickness of 0.4 mm (\lambda\circ/9.7). The developed structures are fabricated with low-cost printed-circuit-board processes on flexible substrates. The transmitarrays are realized by cascading three patterned metallic surfaces (sheet admittances) to achieve complete phase control, while maintaining high transmission. Polarization conversion is accomplished with anisotropic sheets that independently control the field polarized along the two orthogonal axes. The structures are analyzed with both circuit- and fields-based approaches. [ABSTRACT FROM AUTHOR]