Search

The article explores knowledge graphs, a knowledge base that uses a graph-structured data model or topology to integrate data. The author suggests the importance of understanding the historical context and background of one's research area, noting that students and junior researchers are not completely aware of the source of ideas, concepts and techniques they command. Other topics discussed include the automation of reasoning, the history of graphical representations of knowledge, and how the work described in the article was funded by the Millennium Science Initiative Program.

A common approach to constructing fuzzy rough sets (FRSs) is using t-norms. Furthermore, establishing multi-granulation fuzzy rough sets (MGFRSs) is also usually undertaken by means of t-norms. However, most of these sets cannot satisfy the property that upper approximations contain lower approximations under fuzzy binary relations, which is imperative for rough set models. The overlap function is a type of aggregate function that is widely used in multi-attribute decision-making (MADM), image processing and other fields. In this paper, two novel types of MGFRS models under n -dimensional overlap functions are established to overcome the shortcomings of existing models, which are then applied to the multi-attribute group decision-making (MAGDM) problem. First, idempotent n -dimensional overlap functions are used to establish optimistic and pessimistic MGFRSs. These new models fully preserve the important properties of the traditional MGFRS model. Second, through the theoretical analysis of MGFRS based on overlap functions (OMGFRS) and in combination with the TOPSIS method, a solution mechanism for the MAGDM method is proposed. Finally, to fully illustrate this decision-making method, an effective example is developed. A comparison with available methods indicates that this approach is more suitable and has wider adaptability, and that is can be used to handle decision-making problems. [ABSTRACT FROM AUTHOR]

A common approach to constructing fuzzy rough sets (FRSs) is using t-norms. Furthermore, establishing multi-granulation fuzzy rough sets (MGFRSs) is also usually undertaken by means of t-norms. However, most of these sets cannot satisfy the property that upper approximations contain lower approximations under fuzzy binary relations, which is imperative for rough set models. The overlap function is a type of aggregate function that is widely used in multi-attribute decision-making (MADM), image processing and other fields. In this paper, two novel types of MGFRS models under n -dimensional overlap functions are established to overcome the shortcomings of existing models, which are then applied to the multi-attribute group decision-making (MAGDM) problem. First, idempotent n -dimensional overlap functions are used to establish optimistic and pessimistic MGFRSs. These new models fully preserve the important properties of the traditional MGFRS model. Second, through the theoretical analysis of MGFRS based on overlap functions (OMGFRS) and in combination with the TOPSIS method, a solution mechanism for the MAGDM method is proposed. Finally, to fully illustrate this decision-making method, an effective example is developed. A comparison with available methods indicates that this approach is more suitable and has wider adaptability, and that is can be used to handle decision-making problems. [ABSTRACT FROM AUTHOR]

A predictive method using physics-informed machine learning (PIML) and large eddy simulation (LES) is developed to capture the transient flow field through microscale porous media (PSPM). An image processing technique extracts the 3D geometry of the internal layers of the mask from 2D microscopy images, and then the fluid flow is first simulated numerically. The subsequently developed PIML method successfully predicts the transient flow patterns inside the porous medium. For the first time, 3D maps of time-dependent pressure, velocity, and vorticity are predicted across the fibrous porous medium. The results show that, compared to conventional computational fluid dynamics, the PIML method can reduce the computational cost by over 20 times. Further, the LES model can replicate the fine fluctuations caused by the flow passage through the porous medium. Therefore, the developed methodology allows for transient flow predictions in highly complex configurations at a substantially reduced cost. The results indicate that the PIML method can reduce the total computational time (including training and prediction) by 22.5 and 20.7 times over the standard numerical simulation, based on speeds of 0.1 and 0.5 m/s, respectively. Several factors including the inherent differences between CPUs and GPUs, algorithms and software, appear to influence this improvement. [ABSTRACT FROM AUTHOR]

The article discusses the use of artificial intelligence and machine learning in image and video processing for military and aerospace environments. Topics discussed include the latest embedded processing technologies, vision systems in military and aerospace, and the importance of open systems standards to provide vendor-agnostic platforms.

• We improved the background model to extract the background features of the object. • A frequency-domain processing method is proposed to predict the objects' position. • A novel fusion framework is introduced for salient object detection. Salient object detection is a fundamental problem in image processing and computer vision. Many saliency detection algorithms based on the background and frequency-domain are used to extract salient object clues. However, the former causes the real object to be submerged in the detected object areas, especially in complex or small object scenes. While the latter will lead to the loss of some object information when detecting large objects. To solve these problems and achieve better object detection results, we propose a fusion framework for salient object detection by fusing background and frequency-domain features. The background features of the image are extracted by an improved background model. This model represents the spatial layout of the image area with respect to the image boundaries. Meanwhile, we present a new frequency-domain processing method to obtain multiscale frequency-domain features and mark the saliency of the object at different scales. Within our framework, inspired by human visual attention, we use the idea of a self-attention mechanism to capture the intrinsic relation between background and multiscale frequency-domain features. In addition, this fusion framework provides a three-dimensional Gaussian convolution kernel, which expands two-dimensional local information to three dimensions for feature fusion, thus producing more accurate salient objects. Experiment results demonstrate that the proposed method consistently outperforms eleven state-of-the-art methods on five challenging and complicated datasets in terms of four evaluation metrics. [ABSTRACT FROM AUTHOR]

In this brief, we consider the problem of descriptors construction for the task of content-based image retrieval using deep neural networks. The idea of neural codes, based on fully connected layers’ activations, is extended by incorporating the information contained in convolutional layers. It is known that the total number of neurons in the convolutional part of the network is large and the majority of them have little influence on the final classification decision. Therefore, in this brief, we propose a novel algorithm that allows us to extract the most significant neuron activations and utilize this information to construct effective descriptors. The descriptors consisting of values taken from both the fully connected and convolutional layers perfectly represent the whole image content. The images retrieved using these descriptors match semantically very well to the query image, and also, they are similar in other secondary image characteristics, such as background, textures, or color distribution. These features of the proposed descriptors are verified experimentally based on the IMAGENET1M dataset using the VGG16 neural network. For comparison, we also test the proposed approach on the ResNet50 network. [ABSTRACT FROM AUTHOR]

Underwater image processing has been shown to exhibit significant potential for exploring underwater environments. It has been applied to a wide variety of fields, such as underwater terrain scanning and autonomous underwater vehicles (AUVs)-driven applications, such as image-based underwater object detection. However, underwater images often suffer from degeneration due to attenuation, color distortion, and noise from artificial lighting sources as well as the effects of possibly low-end optical imaging devices. Thus, object detection performance would be degraded accordingly. To tackle this problem, in this article, a lightweight deep underwater object detection network is proposed. The key is to present a deep model for jointly learning color conversion and object detection for underwater images. The image color conversion module aims at transforming color images to the corresponding grayscale images to solve the problem of underwater color absorption to enhance the object detection performance with lower computational complexity. The presented experimental results with our implementation on the Raspberry pi platform have justified the effectiveness of the proposed lightweight jointly learning model for underwater object detection compared with the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

The performance of nanoscale semiconductor technologies has become susceptible to high temperatures and aging phenomena. While guard-bands have conventionally been used to combat degradation-induced timing violations, approximations have recently been leveraged to compensate for degradations in lieu of adding timing guard-bands, without a loss in performance. However, only simple approximation techniques such as truncation have been considered in prior work. In this paper, a wide range of approximate arithmetic circuits including adders and multipliers using various sophisticated approximation techniques are investigated to cope with aging- and temperature-induced degradations. To this end, approximate circuits are first characterized for their delay increase under degradations. With this, we then determine the approximation level required to compensate for guard-bands under different degradations. Degradation-aware logic synthesis results show that the simple use of truncated arithmetic circuits leads to a higher quality loss compared to using other approximate circuits. However, a truncated multiplier has the lowest error distance towards a reliable operation in 10 years. The approximate multipliers with configurable error recovery are most suitable when the level of degradation is higher, e.g., at a temperature of 70 °C. The characterization of degradation at the circuit level is then used for design exploration at the architecture level without the need for further gate-level simulations. For three different image processing applications, experimental results show that guard-bands can be mitigated while maintaining an output result with a high visual quality. [ABSTRACT FROM AUTHOR]

• A novel end-to-end framework is designed for satellite super resolution. • A local linear embedding block is proposed to embed sparse patches into the high-level feature space. • A hierarchical non-local block is proposed to facilitate weight learning. • Experiments demonstrate the superiority of the proposed method. Recent years have seen great progress in many deep learning-based super-resolution methods that learn global responses through the non-local strategy. However, improper linear fitting tends to produce unsatisfactory results with notable artifacts. To relieve the improper linear fitting and reuse the hierarchical features, we propose an end-to-end locally linear embedding super-resolution network, termed as LLE-Net, based on the assumption that the local geometric relationship in the low-resolution manifold space also exists in high-level feature manifold space. The proposed LLE-Net includes two major components: 1) the local linear embedding block (LLEB) and 2) the hierarchical non-local block (HNB). In particular, LLEB searches sparse and similar feature maps and embeds the geometric relationship into the high-level feature space, allowing more attention to be paid to objects' textural details. To promote the representation ability, the proposed HNB is able to explore layer- and pixel-level interdependencies. We conduct extensive experiments to evaluate the superiority of LLE-Net via two groups of experiments: (1) super-resolution tasks on two satellite image datasets and a satellite video image dataset, and (2) the subsequent high-level image processing tasks (i.e., satellite semantic segmentation). The proposed method achieves an 0.7 dB improvement in PSNR value on the Draper dataset and 0.07% segmentation accuracy improvement. The source code of the proposed LLE-Net is publicly available. [ABSTRACT FROM AUTHOR]

Enhancing the quality of low-light (LOL) images plays a very important role in many image processing and multimedia applications. In recent years, a variety of deep learning techniques have been developed to address this challenging task. A typical framework is to simultaneously estimate the illumination and reflectance, but they disregard the scene-level contextual information encapsulated in feature spaces, causing many unfavorable outcomes, e.g., details loss, color unsaturation, and artifacts. To address these issues, we develop a new context-sensitive decomposition network (CSDNet) architecture to exploit the scene-level contextual dependencies on spatial scales. More concretely, we build a two-stream estimation mechanism including reflectance and illumination estimation network. We design a novel context-sensitive decomposition connection to bridge the two-stream mechanism by incorporating the physical principle. The spatially varying illumination guidance is further constructed for achieving the edge-aware smoothness property of the illumination component. According to different training patterns, we construct CSDNet (paired supervision) and context-sensitive decomposition generative adversarial network (CSDGAN) (unpaired supervision) to fully evaluate our designed architecture. We test our method on seven testing benchmarks [including massachusetts institute of technology (MIT)-Adobe FiveK, LOL, ExDark, and naturalness preserved enhancement (NPE)] to conduct plenty of analytical and evaluated experiments. Thanks to our designed context-sensitive decomposition connection, we successfully realized excellent enhanced results (with sufficient details, vivid colors, and few noises), which fully indicates our superiority against existing state-of-the-art approaches. Finally, considering the practical needs for high efficiency, we develop a lightweight CSDNet (named LiteCSDNet) by reducing the number of channels. Furthermore, by sharing an encoder for these two components, we obtain a more lightweight version (SLiteCSDNet for short). SLiteCSDNet just contains 0.0301M parameters but achieves the almost same performance as CSDNet. Code is available at https://github.com/KarelZhang/CSDNet-CSDGAN. [ABSTRACT FROM AUTHOR]

This 4K TV builds on last year's EISA Award-winning C825, with improved wide-angle viewing and anti-reflection abilities adding to an enhanced brilliance in picture performance. Primare A35.8 Designed for both the multichannel audio and hi-fi worlds, Primare's flagship 35 Series amplifier deploys eight bridgeable channels of potent, dynamic Hypex Ncore amplification, and then lets the user decide what to do with them. EISA, or the Expert Imaging and Sound Association, is an organization representing 60 of the most respected special interest publications and websites from 29 countries that cover Hi-Fi, Home Theater Video, Home Theater Audio, Photography, Mobile Devices, and In-Car Electronics. [Extracted from the article]

Recently, some authors studied the distributive laws of continuous t-norms and some families of the common classes of uninorms over overlap and grouping functions in [33] , [41] , but until now a complete characterization of the distributivity on idempotent uninorms over overlap or grouping functions widely used in image processing is still unresolved. Moreover, authors in [55] characterized the distributivity equations of uni-nullnorms with continuous Archimedean underlying operators over the above two functions. In this paper, we proceed with the distributivity characterization of idempotent uni-nullnorms over them which obviously generalizes the corresponding results of idempotent uninorms over these two functions. During the process, we introduce a class of weak overlap and grouping functions with weak coefficients, and obtain the full characterizations of the above functions by considering the different values of the underlying uninorms' associated functions of idempotent uni-nullnorms on the interval endpoints and comparing the values with the weak coefficients. These obtained results totally answer the question on the distributive solutions of idempotent uninorms over overlap functions, which have been mentioned as their future works in [33]. In additions, we also obtain that overlap and grouping functions have particular structures with a constant domain where its value equals to the neutral element of the idempotent uninorm when they are distributive over idempotent uninorms. [ABSTRACT FROM AUTHOR]

Image enhancement techniques are commonly used to improve problems such as lack of brightness, high noise, and low contrast in low-light images. Notably, deep learning-based approaches have recently achieved substantial advancements in this domain. However, learning-based methods often require many parameters and multi-layer network structures to achieve high-quality enhancement effects, which limits their application in real-time image processing. To solve this problem, a lightweight Feature Activation Guided Multi-Receptive Field Attention Network (FAMANet) is designed in this paper. The Wavelet Feature Activation Block (WFAB) introduced in the network utilizes the discrete wavelet transform and residual connection to achieve selective activation of image features, thus reducing the redundant information in the feature map and improving the computational efficiency. In addition, the Multi-Receptive Field Attention (MRFA) introduced in this paper addresses the issue of inadequate pixel information and feature map loss stemming from a single input image by concentrating on the image structure, spanning from intricate details to the overall composition. By better-utilizing image information and distinguishing between global and local features, MRFA can improve the speed and efficiency of real-time image processing. After sufficient experimental validation, FAMANet significantly outperforms state-of-the-art methods in low-light image enhancement and exposure correction tasks. [ABSTRACT FROM AUTHOR]

Palmprint recognition techniques have been widely applied in security authentication. As a typical image processing method, convolutional neural network (CNN) has been applied to extract features from palmprint images. However, most existing CNN methods cannot meet the demand in recognition speed due to their high computational complexity. This paper proposes a simple yet effective network (SYEnet) for palmprint recognition, which is a lightweight neural network. SYEnet is composed of four learnable feature extraction networks, which can extract discriminative information from 9 patches. In addition, a dedicated network for palmprint ROI extraction (PREnet) is designed. PREnet extracts ROI from the original image, generating ROI images of consistent dimensions, which can be adapted to complex environments without limitations of lighting, gesture, image quality, and resolution. Moreover, a new contactless palmprint database is collected to validate our method. The database contains 12,800 images from 400 palms and provides data supplementation for current research on palmprint recognition. The proposed SYEnet is evaluated on eight real-world palmprint databases, in comparison to other related methods. Experimental results prove that our method is more efficient and has higher recognition accuracy. [ABSTRACT FROM AUTHOR]

Overlap functions, which can be characterized as a type of non-associative binary aggregation operators, have emerged as one of the most extensively utilized aggregation operators in numerous applications, including image processing, information fusion, and classification problems. At the same time, fuzzy rough sets have also been widely used in these fields due to their excellent ability to handle continuous and uncertain information. However, the variable precision fuzzy rough set model based on overlap functions and its applications have not been fully studied. For instance, some basic properties are invalid and there is a lack of practical applications. In this paper, both overlap functions and precision parameters are introduced into the fuzzy rough sets, namely the overlap function-based variable precision fuzzy rough set (OVPFRS), which is then used in the practical problem of tumor classification. First, considering the existing overlap function-based rough set models, the OVPFRS model is established, and some underlying properties of this model are explored. Second, on the basis of the proposed model, a method for attribute reduction is developed. Finally, the new method is applied to the classification of tumor data from the real world. Through experimentation and comparison with other attribute reduction methods, it has been demonstrated that our model is flexible and the algorithm is viable and effective. [ABSTRACT FROM AUTHOR]

Writing high-performance code on modern machines requires not just locally optimizing inner loops, but globally reorganizing computations to exploit parallelism and locality--doing things such as tiling and blocking whole pipelines to fit in cache. This is especially true for image processing pipelines, where individual stages do much too little work to amortize the cost of loading and storing results to and from off-chip memory. As a result, the performance difference between a naïve implementation of a pipeline and one globally optimized for parallelism and locality is often an order of magnitude. However, using existing programming tools, writing high-performance image processing code requires sacrificing simplicity, portability, and modularity. We argue that this is because traditional programming models conflate the computations defining the algorithm with decisions about intermediate storage and the order of computation, which we call the schedule. We propose a new programming language for image processing pipelines, called Halide, that separates the algorithm from its schedule. Programmers can change the schedule to express many possible organizations of a single algorithm. The Halide compiler then synthesizes a globally combined loop nest for an entire algorithm, given a schedule. Halide models a space of schedules which is expressive enough to describe organizations that match or outperform state-of-the-art hand-written implementations of many computational photography and computer vision algorithms. Its model is simple enough to do so often in only a few lines of code, and small changes generate efficient implementations for x86, ARM, Graphics Processors (GPUs), and specialized image processors, all from a single algorithm. Halide has been public and open source for over four years, during which it has been used by hundreds of programmers to deploy code to tens of thousands of servers and hundreds of millions of phones, processing billions of images every day. [ABSTRACT FROM AUTHOR]

Approximate computing (AC) has been considered as a promising paradigm to improve the energy-efficiency of computing hardware for error-tolerant applications, with negligible quality degradation to the output. Dividers frequently limit the performance of a computing system; however, they have not received as much attention as multipliers and adders in AC. In this paper, an energy-efficient and high-performance approximate divider is proposed based on logarithmic conversion and piecewise constant approximation. In this design, the range for the conversion between binary and logarithmic numbers is first expanded from $\mathbf {[{0,1}]}$ to $\mathbf {[-0.5,1]}$. A heuristic search algorithm is then devised to find the most accurate constant set to approximate the reciprocal of the divisor, by minimizing a statistical error. The hardware implementation is presented for both floating-point (FP) and integer dividers. With a high configurability, the proposed divider results in a mean relative error distance (MRED) from 2.78% to 0.046%, indicating a high accuracy among state-of-the-art approximate dividers. Compared to the half-precision FP divider, the proposed divider with a MRED of 0.74% can achieve nearly $\mathbf {90\times }$ improvement in PDP. Moreover, compared to state-of-the-art approximate dividers, the proposed design is in the Pareto Frontier in terms of power delay product (PDP) and MRED. The three image processing application results demonstrate that the proposed divider can result in the highest peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) even with truncation. [ABSTRACT FROM AUTHOR]

Approximate multipliers are used in error-tolerant applications, sacrificing the accuracy of results to minimize power or delay. In this paper we investigate approximate multipliers using static segmentation. In these circuits a set of $m$ contiguous bits (a segment of $m$ bits) is extracted from each of the two $n$ -bits operand, the two segments are in input to a small $m\times m$ internal multiplier whose output is suitably shifted to obtain the result. We investigate both signed and unsigned multipliers, and for the latter we propose a new segmentation approach. We also present simple and effective correction techniques that can significantly reduce the approximation error with reduced hardware costs. We perform a detailed comparison with previously proposed approximate multipliers, considering a hardware implementation in 28 nm technology. The comparison shows that static segmented multipliers with the proposed correction technique have the desirable characteristic of being on (or close to) the Pareto-optimal frontier for both power vs normalized mean error distance and power vs mean relative error distance trade-off plots. These multipliers, therefore, are promising candidates for applications where their error performance is acceptable. This is confirmed by the results obtained for image processing and image classification applications. [ABSTRACT FROM AUTHOR]

A chaotic map generally employed to generate chaotic sequence is key element for an image encryption algorithm (IEA). In this study, an IEA using novel 2D chaotic map, which is based on Euler and Pi numbers so-called eπ -map, is presented. eπ -map exploits infinity diversity attribute of these numbers. Moreover, a diffusion operation referred to as "bit reversion" in which the bits of the pixels are symmetrically reverted is proposed for manipulating the pixel value. eπ -map is exhaustively examined through bifurcation and trajectory diagrams, Lyapunov exponent (LE), sample entropy (SE), permutation entropy (PE) and 0-1 test. The encryption performance of the IEA is then investigated across various cryptanalysis such as key-space, key sensitivity, histogram, information entropy, correlation coefficient, differential attack, cropping attack, noise attack and encryption execution time. Furthermore, the results are compared with the most recent literature. It is demonstrated that eπ -map has superior hyperchaotic performance in views of ergodicity, complexity and randomness. The IEA based on eπ -map and bit reversion is a secure and reliable algorithm thanks to its outperforming cryptanalysis results. [ABSTRACT FROM AUTHOR]

This paper proposes an analytical design procedure for a class of 2D filters with a directional characteristic in the frequency plane. The method starts from two types of low-pass prototypes, namely a Gaussian filter and a wide-band filter. The two prototypes and the resulted 2D filters are zero-phase, a very useful property in image processing tasks. The prototypes are parametric and can be scaled along the frequency axis, thus resulting narrower or wider. A 1D to 2D frequency mapping, based on accurate approximations, is derived and applied to each prototype, obtaining the desired directional 2D frequency response with specified selectivity and orientation. The designed filters are adjustable, with a frequency response in factored matrix form depending on specifications. The Gaussian prototype yields very selective directional filters, used in detecting oriented straight lines in images, as show simulation results on several test images. Using oriented wide-band filters, other useful 2D filters result, namely square and parallelogram filters. The designed filters have accurate frequency responses, good directional selectivity and high linearity at relatively low order. An advantage of this analytical approach is that the filter transfer function results in parametric form and allows for easy tunability. [ABSTRACT FROM AUTHOR]

A multi-purpose mechanical platform to track moving objects in three-dimensional space has been developed. It is composed of one main microcontroller board that processes all system data, two cameras, three motors, and one secondary microcontroller board to position a platform with three degrees of freedom. The system computes the optical flow and moves the cameras accordingly, tracking motion within the visual scene. The platform operates autonomously. To the best of our knowledge, there are no similar systems reported with low-resolution image sensors and low-cost microcontrollers. Existing solutions rely on personal computers and advanced FPGAs to process image data. This article concludes that the optical flow operation is efficient even using an image sensor with very low resolution. Thus, the system complexity and image data processing are alleviated significantly. The platform can be easily adapted to different application scenarios by adding new peripherals, sensors, or image processing algorithms. A detailed description of the system design and experimental results are provided. [ABSTRACT FROM AUTHOR]

The article reports that the U.S. Department of Defense (DOD) has a bevy of technological hardware and software at its disposal to provide warfighters with the most up-to-date intelligence. Topics include information on the ability to process images captured by a range of devices; and capture details on high-performance embedded computing image and video processors rely heavily on general-purpose graphics processing units (GPGPUs).

Time-series data are widespread in real-world industrial scenarios. To recover and infer missing information in real-world applications, the problem of time-series prediction has been widely studied as a classical research topic in data mining. Deep learning architectures have been viewed as next-generation time-series prediction models. However, recent studies have shown that deep learning models are vulnerable to adversarial attacks. In this study, we prospectively examine the problem of time-series prediction adversarial attacks and propose an attack strategy for generating an adversarial time series by adding malicious perturbations to the original time series to deteriorate the performance of time-series prediction models. Specifically, a perturbation-based adversarial example generation algorithm is proposed using the gradient information of the prediction model. In practice, unlike the imperceptibility to humans in the field of image processing, time-series data are more sensitive to abnormal perturbations and there are more stringent requirements regarding the amount of perturbations. To address this challenge, we craft an adversarial time series based on the importance measurement to slightly perturb the original data. Based on comprehensive experiments conducted on real-world time-series datasets, we verify that the proposed adversarial attack methods not only effectively fool the target time-series prediction model LSTNet, they also attack state-of-the-art CNN-, RNN-, and MHANET-based models. Meanwhile, the results show that the proposed methods achieve a good transferability. That is, the adversarial examples generated for a specific prediction model can significantly affect the performance of the other methods. Moreover, through a comparison with existing adversarial attack approaches, we can see that much smaller perturbations are sufficient for the proposed importance-measurement based adversarial attack method. The methods described in this paper are significant in understanding the impact of adversarial attacks on a time-series prediction and promoting the robustness of such prediction technologies. [ABSTRACT FROM AUTHOR]

Recently, single-particle cryo-electron microscopy (cryo-EM) has become an indispensable method for determining macromolecular structures at high resolution to deeply explore the relevant molecular mechanism. Its recent breakthrough is mainly because of the rapid advances in hardware and image processing algorithms, especially machine learning. As an essential support of single-particle cryo-EM, machine learning has powered many aspects of structure determination and greatly promoted its development. In this article, we provide a systematic review of the applications of machine learning in this field. Our review begins with a brief introduction of single-particle cryo-EM, followed by the specific tasks and challenges of its image processing. Then, focusing on the workflow of structure determination, we describe relevant machine learning algorithms and applications at different steps, including particle picking, 2-D clustering, 3-D reconstruction, and other steps. As different tasks exhibit distinct characteristics, we introduce the evaluation metrics for each task and summarize their dynamics of technology development. Finally, we discuss the open issues and potential trends in this promising field. [ABSTRACT FROM AUTHOR]

Fuzzy inference with the Bandler-Kohout subproduct (BKS) has been successfully applied in many fields such as fuzzy control, artificial intelligence, image processing, data mining, decision-making, prediction, classification and so on. However, one has to face with the rule explosion in these applications. To deal with this problem, hierarchical fuzzy systems with the compositional rule of inference (CRI) method have been constructed by a series of low-dimensional sub fuzzy systems. And it has been proved that hierarchical fuzzy inference method can efficiently restrain the explosion of fuzzy rules. Therefore, in order to increase the computational efficiency of the fuzzy inference based on the BKS when multi-input-single-output (MISO) fuzzy rules are involved, this paper mainly constructs two hierarchical fuzzy inference methods based on the BKS in which the if-then rules are respectively interpreted by fuzzy implications and ML -implications. Moreover, the validity of the two BKS hierarchical fuzzy inferences is studied with the GMP rules. Finally, two examples are employed to illustrate the computational efficiency of our proposed BKS hierarchical inference methods. [ABSTRACT FROM AUTHOR]

Glioblastoma is the most common and deadly primary brain tumor in adults. To optimize the treatment strategies, it is essential to understand the tumor growth dynamics in different periods. In this study, we use image processing techniques to combine the available early-stage imaging data and applied a fractional reaction–diffusion equation to predict the human glioblastoma multiforme tumor growth. We consider the heterogeneity of the brain tissue by assigning different diffusion coefficients for the three regions of the human brain. A meshfree method based on the thin plate spline radial basis function is used for the numerical solution of nonlinear time fractional Proliferation-Invasion equation. The results show that the proposed model has a better fit with the experimental data. The prediction of tumor growth at any desired time with no need to repeated imaging is another advantages of the model which could reduce the side effects and cost of diagnostic and therapeutic methods. The model can also incorporate the effects of various treatment modalities such as hyperthermia, radiation, and surgery on tumor growth. Furthermore, it can enable the use of patient-specific characteristics in diagnosis and treatment and facilitate the development of personalized medicine approaches. [ABSTRACT FROM AUTHOR]

Overlap functions are an important class of aggregation operators on [0,1] that have been proposed for applications in image processing, classification, etc. Later, Paiva et al. lifted overlap functions on [0,1] to complete lattices. In this paper, we continue to study overlap functions on bounded posets so as to lift the continuity in the notion of overlap functions from [0,1] to bounded posets mainly from the topological aspects. More precisely, we introduce the notion of overlap functions on bounded posets. And then, we consider the multiplicative generator triples of overlap functions on bounded posets, and investigate constructions of overlap functions by multiplicative generator triples. [ABSTRACT FROM AUTHOR]

This research aims to revolutionize complex image clipping using quantum computing, particularly in line clipping algorithms. Traditional methods like Cohen-Sutherland, Liang-Barsky, Cyrus-Beck, and Fast Clipping have limitations in scalability and efficiency. The proposed quantum line clipping algorithm uses quantum computing's unique properties to redefine image processing. It views images as vectors, allowing for computation of pixel distances through quantum inner products. The introduction of Q-AND and Q-OR gates within a nine-section quantum display unit plays a pivotal role in decision-making for line clipping, enabling the exclusion of lines beyond the viewport. Comparing this algorithm with classical methods, the quantum algorithm on IBM, Google, and Rigetti's platforms demonstrated superior runtime complexities, demonstrating the immense potential of quantum computation in addressing intricate image clipping challenges. The Quantum Fourier Transform (QFT) acts as a quantum black box, generating high probability amplitudes for substantial vectors. The research also introduces mathematical theorems, such as the quantum image clipping equivalence theorem, which establishes a robust equivalence relation between quantum pixels in frame buffers, the quantum way line clipping equivalence theorem, and the quantum support line clipping optimality theorem. The practical implementation of the proposed quantum line clipping algorithm is presented in Appendix-B, providing valuable resources for researchers and practitioners seeking to explore and experiment with quantum line clipping techniques. [ABSTRACT FROM AUTHOR]

Face-attribute synthesis is a typical application of neural network technology. However, most current methods suffer from the problem of uncontrollable attribute intensity. In this study, we proposed a novel intensity-controllable generation network (ICGNet) based on covering learning for face attribute synthesis. Specifically, it includes an encoder module based on the principle of homology continuity between homologous samples to map different facial images onto the face feature space, which constructs sufficient and effective representation vectors by extracting the input information from different condition spaces. It then models the relationships between attribute instances and representational vectors in space to ensure accurate synthesis of the target attribute and complete preservation of the irrelevant region. Finally, the progressive changes in the facial attributes by applying different intensity constraints to the representation vectors. ICGNet achieves intensity-controllable face editing compared to other methods by extracting sufficient and effective representation features, exploring and transferring attribute relationships, and maintaining identity information. The source code is available at https://github.com/kllaodong/-ICGNet. • We designed a new encoder module to map face images of different condition spaces into face feature space to obtain sufficient and effective face feature representation. • Based on feature extraction, we proposed a novel Intensity-Controllable Generation Network (ICGNet), which can realize face attribute synthesis with continuous intensity control while maintaining identity and semantic information. • The quantitative and qualitative results showed that the performance of ICGNet is superior to current advanced models. • New encoder module maps varied face images into a unified feature space. • ICGNet innovatively synthesizes face attributes with intensity control. • ICGNet maintains identity and semantics in face attribute generation. • Outperforms advanced models in quantitative and qualitative measures. [ABSTRACT FROM AUTHOR]

Currently, weapon detection through video surveillance has been extensively studied using deep learning techniques by researchers. However, limited attention has been given to the detection of weapons in night time or dark scenarios. This paper aims to address this gap by proposing a novel approach for weapon detection specifically modified for low-light conditions. The authors demonstrate accurate and robust detection of weapons in challenging nighttime environments by modifying the deep learning model YOLOv7. The YOLOv7-DarkVision model has been developed by combining a brightening algorithm with the advanced image processing techniques and architecture of YOLOv7. A dataset of 15,367 images were collected for training of the model, along with five dark videos from various sources for performance evaluation. The derived detection model, which has a precision score of 95.50% and an F1-Score of 93.41%, performs absolutely well as a weapon detector. [ABSTRACT FROM AUTHOR]

In recent years, overlap functions, as a class of bivariate aggregation operators that are widely used in various application problems (see, e.g., in decision-making, image processing, classifications etc.), have been generalized to many forms. In particular, Paiva et al. (R. Paiva, E. Palmeira, R. Santiago, B. Bedregal, Lattice-valued overlap and quasi-overlap functions, Information Sciences 562 (2021) 180–199.) have generalized the overlap functions to the so-called quasi-overlap functions lately. In the meantime, considering aggregation operators on finite chains, especially the commonly bivariate aggregation operators (see, e.g., t-norms, t-conorms, uninorms, t-operators etc.) has become an important research topic in the fields of aggregation operators. In this paper, we take into account this research topic for quasi-overlap functions. First of all, we give the concept of quasi-overlap functions on a finite chain L with n + 2 elements and its arbitrary subchains together with three generalized forms of quasi-overlap functions on any subchain of L. And then, we show some examples of quasi-overlap functions on L along with some of its specific subchains and study the idempotent property, Archimedean property and cancellation law of quasi-overlap functions on L. Finally, we obtain two construction methods of quasi-overlap functions on L , one of them is the ordinal sum construction. [ABSTRACT FROM AUTHOR]

Recently, integrating ferroelectric materials with nanotransistors such as carbon nanotube field-effect transistors (CNTFETs) has opened new doors for demonstrating a new generation of ultra-miniature circuits and systems. Utilizing the negative differential resistance effect in negative capacitance CNTFETs (NC-CNTFETs) has spurred the efforts for designing ultra-compact ternary circuits and systems similar to their binary structures. This paper presents an ultra-efficient ternary image edge detection hardware using NC-CNTFET technology. The proposed hardware is endowed with a noise reduction circuitry to mitigate the noise effects. Using four $1\times 3$ kernels and concatenating the image pixels, the proposed ternary hardware has been designed using only 50 transistors. The proposed ternary hardware at the circuit level shows, on average, 74% improvements regarding power delay-product (PDP) compared to the CNTFET-based counterparts. Our comprehensive simulations indicate that the proposed NC-CNTFET-based hardware shows a 40% improvement in data loss, 2.2 times improvement in performance ratio, and 1.14 times improvement in Pratt’s figure-of-merit, respectively, compared to the related designs. Our results accentuate that the proposed NC-CNTFET-based ternary hardware is a breakthrough achievement in demonstrating ultra-efficient and noise-immune ternary image processing circuits beyond the conventional binary counterparts. [ABSTRACT FROM AUTHOR]

It is known that the human visual system performs a hierarchical information process in which early vision cues (or primitives) are fused in the visual cortex to compose complex shapes and descriptors. While different aspects of the process have been extensively studied, such as lens adaptation or feature detection, some other aspects, such as feature fusion, have been mostly left aside. In this work, we elaborate on the fusion of early vision primitives using generalizations of the Choquet integral, and novel aggregation operators that have been extensively studied in recent years. We propose to use generalizations of the Choquet integral to sensibly fuse elementary edge cues, in an attempt to model the behaviour of neurons in the early visual cortex. Our proposal leads to a fully-framed edge detection algorithm whose performance is put to the test in state-of-the-art edge detection datasets. [ABSTRACT FROM AUTHOR]

Three-dimensional (3D) face application has attracted significant attention the field of multimedia and image processing. However, the end-to-end 3D face reconstruction method is still immature, and there are some problems, such as overfitting caused by too few training sets and unacceptable 3D face texture alignment performance. Therefore, we design a novel approach to construct a 3D face, named multi-objective evolutionary 3D face reconstruction based on improved encoder–decoder network (MoEDN). This study introduces a regularization algorithm named feature map distortion (Disout); whose purpose is to strengthen the network generalization ability. Based on this, we construct a multi-objective evolutionary 3D face reconstruction model, in which decision variables are distortion probability, distorted block size, distorted intensity, probability step, and learning rate; and objective functions are loss and structural similarity (SSIM). We use four multi-objective evolutionary algorithms (NSGA-II, AGEII, NSLS, and MOEA/D) to optimize the proposed model. Experimental results demonstrate that NSLS has the best performance. In addition, compared with position map regression network (PRNet), 2D-assisted self-supervised learning (2DASL) and other state-of-the-art, the proposed model achieves better loss values and NME values. Therefore, the proposed multi-objective evolutionary 3D face reconstruction model has outstanding 3D facial reconstruction performance in large poses and face expression. [ABSTRACT FROM AUTHOR]

In the last years fuzzy (quasi-) metrics and indistinguishability operators have been used as a mathematical tool in order to develop appropriate models useful in applied sciences as, for instance, image processing, clustering analysis and multi-criteria decision making. The both aforesaid similarities allow us to fuzzify the crisp notion of equivalence relation when a certain degree of similarity can be only provided between the compared objects. However, the applicability of fuzzy (quasi-) metrics is reduced because the difficulty of generating examples. One technique to generate new fuzzy binary relations is based on merging a collection of them into a new one by means of the use of a function. Inspired, in part, by the preceding fact, this paper is devoted to study which functions allow us to merge a collection of fuzzy (quasi-) metrics into a single one. We present a characterization of such functions in terms of ∗ -triangular triplets and also in terms of isotonicity and ∗ -supmultiplicativity, where ∗ is a t-norm. We also show that this characterization does not depend on the symmetry of the fuzzy quasi-metrics. The same problem for stationary fuzzy (quasi-) metrics is studied and, as a consequence, characterizations of those functions aggregating fuzzy preorders and indistinguishability operators are obtained. [ABSTRACT FROM AUTHOR]

Transistor aging phenomena manifest themselves as degradations in the main electrical characteristics of transistors. Over time, they result in a significant increase of cell propagation delay, leading to errors due to timing violations, since the operating frequency becomes unsustainable as the circuit ages. Conventional techniques employ timing guardbands to mitigate aging-induced delay increase, which leads to considerable performance losses from the beginning of the circuit’s lifetime. Leveraging the inherent error resilience of a vast number of application domains, approximate computing was recently introduced as an aging mitigation mechanism. In this work, we present the first automated framework for generating aging-aware approximate circuits. Our framework, by applying directed gate-level netlist approximation, induces a small functional error and recovers the delay degradation due to aging. As a result, our optimized circuits eliminate aging-induced timing errors. Experimental evaluation over a variety of arithmetic circuits and image processing benchmarks demonstrates that for an average error of merely $5\times 10^{-3}$ , our framework completely eliminates aging-induced timing guardbands. Compared to the respective baseline circuits without timing guardbands (i.e., iso-performance evaluation), the error of the circuits generated by our framework is $1208\times $ smaller. [ABSTRACT FROM AUTHOR]

Image processing with any potential quantum computing hardware requires a quantum color model capable of capturing and manipulating color information in images. In this study, a quantum hue, saturation, and lightness (QHSL) model is proposed as a first attempt to encode perceptually relevant triplet color components using the properties of quantum mechanics (i.e., entanglement and parallelism). The proposed color model was used to define a representation of two-dimensional QHSL images for storage and transformation with fewer computing resources. The configuration of color-assignment attributes within this QHSL representation offer useful applications for image analysis. Specifically, a pseudocolor technique with flexible gray depth divisions is presented for highlighting fine visual details. [ABSTRACT FROM AUTHOR]

Approximate computing is an efficient approach for reducing computational costs. This method involves a trade-off between computational accuracy and the circuit’s power consumption, delay, and area. However, the accuracy requirements may differ for different applications. In some situations, precise results are required. Therefore, this paper proposes an accuracy-configurable radix-4 adder (ACRA), which uses the power gating technique to turn on or turn off the partial logic gates of an adder element dynamically to compute accurate or approximate results. When the ACRA operates in the approximate mode, the partial sum of one adder element is modified to reduce the error distance between the approximate and accurate results. A comparison of the ACRA with two state-of-the-art accuracy-configurable adders indicated that the ACRA achieved the best trade-off between the power–delay product and computational accuracy. In addition, in the image processing experiment conducted in this study, under half-approximate computation, the peak signal-to-noise ratio of the ACRA was less than 1 dB and its structural similarity index measure was maintained above 0.99. These results are superior to those obtained for other accuracy-configurable adders. [ABSTRACT FROM AUTHOR]

As an important class of aggregation operators, the notion of overlap functions was first presented in 2009 in order to be considered for applications in image processing context. Later, many other researches arised bringing some variations of those functions for different purposes. Here, our main goal is defining overlap functions on lattices and discuss how a weakned version of it, named quasi-overlaps, works when continuity is eliminated from the definition. Some properties of quasi-overlaps on lattices, namely convex sum, migrativity, homogeneity, idempotency and cancellation law are investigated. Also, Finally, properties related to continuity as "Archimedean" and "limiting" are studied. [ABSTRACT FROM AUTHOR]

Currently, the demand for covert communication in open social networks brings new opportunities and challenges to existing image steganography technology in terms of robustness and security. To this end, an image robust adaptive steganography is proposed with robustness against multiple image processing attacks and detection resistance. First, a robust embedding domain with theoretical foundation and optimal invisibility is constructed based on the compression resistance principle. Then, utilizing the robust image abstraction and saliency measurement, the embedding channel is selected to avoid modifications in smooth regions and enhance visual quality. On this basis, the proposed method is given combining with error-correcting and STC codes to realize message embedding with minimum costs and improve extraction accuracy. Lastly, after parameters discussion and selection, the performance experiments are conducted compared with previous representative steganography algorithms, concerning robustness and detection resistance, and the fault tolerance is deduced, thereby providing the recommended coding parameters to improve message extraction integrity. The experimental results show that the proposed method can realize message extraction with high accuracy after JPEG compression, Gaussian noising, and scaling attacks, while holding comparable detection resistance to adaptive steganography against statistical features, which indicates its application prospect for covert communication in open lossy channels. [ABSTRACT FROM AUTHOR]

We present an approximate logarithmic multiplier with two-stage operand trimming, which prioritises area and energy consumption while retains acceptable accuracy. The multiplier trims the least significant parts of input operands in the first stage and the mantissas of the obtained operands’ approximations in the second stage. We evaluated the multiplier’s efficiency in terms of error, energy, and area utilisation using NanGate 45nm. The experimental results show that the proposed multiplier exhibits smaller area utilisation and energy consumption than the state-of-the-art designs and that it behaves well in image processing and image classification with convolutional neural networks. [ABSTRACT FROM AUTHOR]

Individual identification of animals from camera traps has become an important task in wildlife research, but camera deployment methods often do not facilitate this important undertaking. Identification of individual golden eagles (Aquila chrysaetos) is possible using uniquely marked rectrices, but no studies have explored methods to maximize the rate of individual identification from camera images. Our objectives were to assess whether different camera heights (1 m vs. 3 m), image capture settings (one image after a 1‐min delay vs. burst of 5 images after a 30 sec delay), and arrangements relative to bait (dorsally vs. ventrally aimed) affected views of rectrices on golden eagles and our ability to identify individuals. We conducted our study from 15 December 2016 to 3 March 2017 on the Savannah River Site, South Carolina. First, we developed a scoring system based on views of rectrices and used a linear mixed‐effects model to compare image scores among different camera arrangements and image settings. Next, after identifying individual eagles, we used generalized linear mixed‐effects models to compare total individual eagle detections, total days an individual was detected, and probability of obtaining an unknown individual identification among camera arrangements and settings. Overall, we scored a total of 27,499 images, with 8,083 providing views of marked rectrices that allowed identification of 18 individual eagles. Average image scores and proportion of images suitable for individual identification were higher from elevated (3 m) camera arrangements than standard arrangements (1 m) across sites. Regardless of camera height, faster frequency of image capture provided more images that could be used to identify individuals and the most trap days per individual. Researchers and managers should consider deploying elevated cameras traps with faster frequency of image capture to improve data quality and potential for analysis of golden eagle populations and trends across the species' range. Published 2021. This article is a U.S. Government work and is in the public domain in the USA. : Elevated cameras using faster image capture settings can provide better and more images of unique golden eagle (Aquila chrysaetos) rectrices, regardless of bait positions and motion sensor idiosyncrasies. Researchers and managers should consider deploying elevated remote cameras with faster frequency of image capture to maximize rate of individual identification when monitoring golden eagle populations and trends across the species' range. [ABSTRACT FROM AUTHOR]

Publishing/sharing pretrained deep neural network (DNN) models is a common practice in the community of computer vision. The increasing popularity of pretrained models has made it a serious concern: how to protect the intellectual properties of model owners and avert illegal usages by malicious attackers. This article aims at developing a framework for watermarking DNNs, with a particular focus on low-level image processing tasks that map images to images. Using image denoising and superresolution as case studies, we develop a black-box watermarking method for pretrained models, which exploits the overparameterization of the DNNs in image processing. In addition, an auxiliary module for visualizing the watermark information is proposed for further verification. Extensive experiments show that the proposed watermarking framework has no noticeable impact on model performance and enjoys the robustness against the often-seen attacks. [ABSTRACT FROM AUTHOR]

General-purpose processors, while tremendously versatile, pay a huge cost for their flexibility by wasting over 99% of the energy in programmability overheads. We observe that reducing this waste requires tuning data storage and compute structures and their connectivity to the data-flow and data-locality patterns in the algorithms. Hence, by backing off from full programmability and instead targeting key data-flow patterns used in a domain, we can create efficient engines that can be programmed and reused across a wide range of applications within that domain. We present the Convolution Engine (CE)--a programmable processor specialized for the convolution-like data-flow prevalent in computational photography, computer vision, and video processing. The CE achieves energy efficiency by capturing data-reuse patterns, eliminating data transfer overheads, and enabling a large number of operations per memory access. We demonstrate that the CE is within a factor of 2–3× of the energy and area efficiency of custom units optimized for a single kernel. The CE improves energy and area efficiency by 8–15× over data-parallel Single Instruction Multiple Data (SIMD) engines for most image processing applications. [ABSTRACT FROM AUTHOR]

The article discusses the pixel-labeling or semantic-segmentation in programming computers to allow automatic interpretation of the image content. Topics discussed include the challenge in artificial intelligence and computer vision, hierarchical segmentation or scene parsing to produce refined view of the scene, the process in pixel labeling, probabilistic model known as conditional Markov random field (CRF) use in most methods for pixel labeling and partition function that ensures the probability distribution.

This paper proposes a technique for processing handwriting images. The algorithm used in this study is an improvement to the binarisation process. The enhancement focuses on correcting damaged lines that usually arise during the binarisation process, particularly, spurious holes, discontinuities, and eroded boundaries. The presented method uses a morphogical dilation operation in which a structural element is locally adapted using the information from a directional field. The adaptation process involves a new criterion for selecting orientation and shape of a structural element that combines directional field, a coherence measure, and a circular histogram. The field was computed using gradient-based approach, and a method based on a Hessian matrix. During experiments, our method was applied to the output of selected binarisation algorithms. The experiments were conducted on grayscale signature images (from the CEDAR database) and handwriting images (from the DIBCO database). The results of the algorithm were compared to the results of standard morphological operations (dilation, erosion, opening, and closing) and median filtering. The experiments show that the proposed method achieves significant accuracy improvement (8%–12% for Acc, 15%–32% for Acc2 measures), reduces the number of unwanted artefacts, and produces images with less distortion compared to those from standard approaches. [ABSTRACT FROM AUTHOR]

This paper proposes a technique for processing handwriting images. The algorithm used in this study is an improvement to the binarisation process. The enhancement focuses on correcting damaged lines that usually arise during the binarisation process, particularly, spurious holes, discontinuities, and eroded boundaries. The presented method uses a morphogical dilation operation in which a structural element is locally adapted using the information from a directional field. The adaptation process involves a new criterion for selecting orientation and shape of a structural element that combines directional field, a coherence measure, and a circular histogram. The field was computed using gradient-based approach, and a method based on a Hessian matrix. During experiments, our method was applied to the output of selected binarisation algorithms. The experiments were conducted on grayscale signature images (from the CEDAR database) and handwriting images (from the DIBCO database). The results of the algorithm were compared to the results of standard morphological operations (dilation, erosion, opening, and closing) and median filtering. The experiments show that the proposed method achieves significant accuracy improvement (8%–12% for Acc, 15%–32% for Acc2 measures), reduces the number of unwanted artefacts, and produces images with less distortion compared to those from standard approaches. [ABSTRACT FROM AUTHOR]

Invisibility is an important target for image steganography; however, the traditional image steganography methods inevitably leave traces of their modifications in cover images due to the process used to embed secret messages. In this paper, we propose an encrypted coverless information hiding method that transfers secret images between two different image domains using generative models. Our method includes two stages: encryption and decryption. In the encryption stage, we first embed a secret image into a public image (one domain) to obtain a synthetic image; then, we utilize that image as the input to the first generative model F to obtain an encrypted image (another domain). Adversarial loss and an extraction module are added to improve the quality of the encrypted images generated in this stage. In the decryption stage, we design a second generative model G to reconstruct the synthetic images from the encrypted images. Finally, the secret image is separated from the reconstructed synthetic image. In our extensive experiments, we adopt the images in the MNIST dataset as the secret images, which allows us to use the recognition rate as a measure of the secret image reconstruction quality. The experimental results indicate that our method is not only able to generate quality encrypted images compared with current popular image-to-image translation methods but also possesses greater security, robustness and reconstruction ability. [ABSTRACT FROM AUTHOR]

Although there exist various adversarial attacking methods, most of them are performed by generating adversarial noises. Inspired by the fact that people usually set different camera parameters to obtain diverse visual styles when taking a picture, we propose the adversarial attributes, which generate adversarial examples by manipulating the image attributes like brightness, contrast, sharpness, chroma to simulate the imaging process. This task is accomplished under the black-box setting, where only the predicted probabilities are known. We formulate this process into an optimization problem. After efficiently solving this problem, the optimal adversarial attributes are obtained with limited queries. To guarantee the realistic effect of adversarial examples, we bound the attribute changes using L p norm versus different p values. Besides, we also give a formal explanation for the adversarial attributes based on the linear nature of Deep Neural Networks (DNNs). Extensive experiments are conducted on two public datasets, including CIFAR-10 and ImageNet with respective to four representative DNNs like VGG16, AlexNet, Inception v3 and Resnet50. The results show that at most 97.79 % of images in CIFAR-10 test dataset and 98.01 % of the ImageNet images can be successfully perturbed to at least one wrong class with only ⩽ 300 queries per image on average. [ABSTRACT FROM AUTHOR]