Search

This document is the closing editorial for a special issue of the journal Applied Sciences on computer vision and pattern recognition based on deep learning. The issue includes 31 papers covering various topics such as image and video processing, object detection, object and scene recognition, visual application technologies, classification, segmentation, compression, and more. Each paper explores different methods and techniques to improve the performance and accuracy of deep learning models in these areas. The document provides a brief summary of each paper, highlighting their contributions and findings. [Extracted from the article]

In this paper, we discuss the implementation of artificial self-awareness mechanisms and self-reflection abilities in deep neural networks. While the current limitations of research prevent achieving cognitive capabilities on par with natural biological entities, the incorporation of basic self-awareness and self-reflection mechanisms in deep learning architectures offers substantial advantages in tackling specific problems across various scientific fields, including geosciences. In the first section, we outline the foundational architecture of our deep learning approach termed Self-Aware Learning (SAL). The subsequent part of the paper highlights the practical benefits of this machine learning methodology through synthetic tests and applications addressed to automatic classification and image analysis of real petrological data sets. We show how Self-Aware Learning allows enhanced accuracy, reduced overfitting problems, and improved performances compared to other existing methods. [ABSTRACT FROM AUTHOR]

Sports image classification has gained increasing significance in various domains, including athlete training, event management, and broadcasting. Nonetheless, accurately classifying these images remains a formidable challenge due to factors such as variations in lighting conditions, pose, and attire. This paper proposes an advanced deep learning approach to address this issue. The proposed method uses a modified Inception-V4 (IV4) architecture, which has exhibited remarkable success in image classification tasks. To further enhance the performance of the IV4 model, an advanced version of the seasons optimizer (ASO), a potent optimization algorithm, has been introduced, which adjusts the learning rate during training. This modified optimizer enables the IV4 model to converge more rapidly and achieve higher accuracy. An evaluation of the proposed method has been conducted using a comprehensive dataset comprising Olympic sports images. The experimental results demonstrate significant enhancements over existing approaches, attaining superior performance across multiple evaluation metrics. These findings underscore the potential of combining cutting-edge deep neural networks with advanced optimization techniques for the precise and efficient classification of sports images. [ABSTRACT FROM AUTHOR]

Learning in neural networks with locally-tuned neuron models such as radial Basis Function (RBF) networks is often seen as instable, in particular when multi-layered architectures are used. Furthermore, universal approximation theorems for single-layered RBF networks are very well established; therefore, deeper architectures are theoretically not required. Consequently, RBFs are mostly used in a single-layered manner. However, deep neural networks have proven their effectiveness on many different tasks. In this paper, we show that deeper RBF architectures with multiple radial basis function layers can be designed together with efficient learning schemes. We introduce an initialization scheme for deep RBF networks based on k-means clustering and covariance estimation. We further show how to make use of convolutions to speed up the calculation of the Mahalanobis distance in a partially connected way, which is similar to the convolutional neural networks (CNNs). Finally, we evaluate our approach on image classification as well as speech emotion recognition tasks. Our results show that deep RBF networks perform very well, with comparable results to other deep neural network types, such as CNNs. [ABSTRACT FROM AUTHOR]

Deep learning (DL) and convolutional neural networks (CNNs) have achieved state-of-the-art performance in many medical image analysis tasks. Histopathological images contain valuable information that can be used to diagnose diseases and create treatment plans. Therefore, the application of DL for the classification of histological images is a rapidly expanding field of research. The popularity of CNNs has led to a rapid growth in the number of works related to CNNs in histopathology. This paper aims to provide a clear overview for better navigation. In this paper, recent DL-based classification studies in histopathology using strongly annotated data have been reviewed. All the works have been categorized from two points of view. First, the studies have been categorized into three groups according to the training approach and model construction: 1. fine-tuning of pre-trained networks for one-stage classification, 2. training networks from scratch for one-stage classification, and 3. multi-stage classification. Second, the papers summarized in this study cover a wide range of applications (e.g., breast, lung, colon, brain, kidney). To help navigate through the studies, the classification of reviewed works into tissue classification, tissue grading, and biomarker identification was used. [ABSTRACT FROM AUTHOR]

The use of deep learning techniques in real-time monitoring can save a lot of manpower in various scenarios. For example, mask-wearing is an effective measure to prevent COVID-19 and other respiratory diseases, especially for vulnerable populations such as children, the elderly, and people with underlying health problems. Currently, many public places such as hospitals, nursing homes, social service facilities, and schools experiencing outbreaks require mandatory mask-wearing. However, most of the terminal devices currently available have very limited GPU capability to run large neural networks. This means that we have to keep the parameter size of a neural network modest while maintaining its performance. In this paper, we propose a framework that applies deep learning techniques to real-time monitoring and uses it for the real-time monitoring of mask-wearing status. The main contributions are as follows: First, a feature fusion technique called skip layer pooling fusion (SLPF) is proposed for image classification tasks. It fully utilizes both deep and shallow features of a convolutional neural network while minimizing the growth in model parameters caused by feature fusion. On average, this technique improves the accuracy of various neural network models by 4.78% and 5.21% on CIFAR100 and Tiny-ImageNet, respectively. Second, layer attention (LA), an attention mechanism tailor-made for feature fusion, is proposed. Since different layers of convolutional neural networks make different impacts on the final prediction results, LA learns a set of weights to better enhance the contribution of important convolutional layer features. On average, it improves the accuracy of various neural network models by 2.10% and 2.63% on CIFAR100 and Tiny-ImageNet, respectively. Third, a MobileNetv2-based lightweight mask-wearing status classification model is trained, which is suitable for deployment on mobile devices and achieves an accuracy of 95.49%. Additionally, a ResNet mask-wearing status classification model is trained, which has a larger model size but achieves high accuracy of 98.14%. By applying the proposed methods to the ResNet mask-wearing status classification model, the accuracy is improved by 1.58%. Fourth, a mask-wearing status detection model is enhanced based on YOLOv5 with a spatial-frequency fusion module resulting in a mAP improvement of 2.20%. Overall, this paper presents various techniques to improve the performance of neural networks and apply them to mask-wearing status monitoring, which can help stop pandemics. [ABSTRACT FROM AUTHOR]

Hyperspectral remote sensing images (HSIs) have both spectral and spatial characteristics. The adept exploitation of these attributes is central to enhancing the classification accuracy of HSIs. In order to effectively utilize spatial and spectral features to classify HSIs, this paper proposes a method for the spatial feature extraction of HSIs based on a mutually guided image filter (muGIF) and combined with the band-distance-grouped principal component. Firstly, aiming at the problem that previously guided image filtering cannot effectively deal with the inconsistent information structure between the guided and target information, a method for extracting spatial features using muGIF is proposed. Then, aiming at the problem of the information loss caused by a single principal component as a guided image in the traditional GIF-based spatial–spectral classification, a spatial feature-extraction framework based on the band-distance-grouped principal component is proposed. The method groups the bands according to the band distance and extracts the principal components of each set of band subsets as the guide map of the current band subset to filter the HSIs. A deep convolutional neural network model and a generative adversarial network model for the filtered HSIs are constructed and then trained using samples for HSIs' spatial–spectral classification. Experiments show that compared with the traditional methods and several popular spatial–spectral HSI classification methods based on a filter, the proposed methods based on muGIF can effectively extract the spatial–spectral features and improve the classification accuracy of HSIs. [ABSTRACT FROM AUTHOR]

This paper addresses the performance degradation of deep neural networks caused by learning with noisy labels. Recent research on this topic has exploited the memorization effect: networks fit data with clean labels during the early stages of learning and eventually memorize data with noisy labels. This property allows for the separation of clean and noisy samples from a loss distribution. In recent years, semi-supervised learning, which divides training data into a set of labeled clean samples and a set of unlabeled noisy samples, has achieved impressive results. However, this strategy has two significant problems: (1) the accuracy of dividing the data into clean and noisy samples depends strongly on the network's performance, and (2) if the divided data are biased towards the unlabeled samples, there are few labeled samples, causing the network to overfit to the labels and leading to a poor generalization performance. To solve these problems, we propose the curriculum regularization and adaptive semi-supervised learning (CRAS) method. Its key ideas are (1) to train the network with robust regularization techniques as a warm-up before dividing the data, and (2) to control the strength of the regularization using loss weights that adaptively respond to data bias, which varies with each split at each training epoch. We evaluated the performance of CRAS on benchmark image classification datasets, CIFAR-10 and CIFAR-100, and real-world datasets, mini-WebVision and Clothing1M. The findings demonstrate that CRAS excels in handling noisy labels, resulting in a superior generalization and robustness to a range of noise rates, compared with the existing method. [ABSTRACT FROM AUTHOR]

Image recognition has long been one of the research hotspots in computer vision tasks. The development of deep learning is rapid in recent years, and convolutional neural networks usually need to be designed with fixed resources. If sufficient resources are available, the model can be scaled up to achieve higher accuracy, for example, VggNet, ResNet, GoogLeNet, etc. Although the accuracy of large-scale models has been improved, the following problems will occur with the expansion of model scale: (1) There may be over-fitting; (2) increasing model parameters; (3) slow model convergence. This paper proposes a design method for a modular convolutional neural network model which solves the problem of over-fitting and large model parameters by connecting multiple modules in parallel. Moreover, each module contains several submodules (three submodules in this paper) and fuses the features extracted from the submodules. The model convergence can be accelerated by using the fused features (the fused features contain more image information). In this study, we add a gate unit based on the attention mechanism to the model, which aims to optimize the structure of the model (select the optimal number of modules), allowing the model to select an optimum network structure by learning and dynamically reducing FLOPs (floating-point operations per second) of the model. Compared to VggNet, ResNet, and GoogLeNet, the structure of the model proposed in this paper is simple and the parameters are small. The proposed model achieves good results in the Kaggle datasets Cats-vs.-Dogs (99.3%), 10-Monkey Species (99.26%), and Birds-400 (99.13%). [ABSTRACT FROM AUTHOR]

In this paper, a microscopic oil displacement visualization experiment based on the glass etching model to simulate the tight oil reservoir of underground rocks is carried out. At present, water flooding technology is widely used in the development of oil and gas fields, and the remaining oil content is still very high after water flooding. It is the key to improving oil recovery to identify and study the remaining oil form distribution after water flooding. The experiment result shows there are five types of residual oil after water flooding: columnar residual oil, membranous residual oil, oil droplet residual oil, blind terminal residual oil and cluster residual oil. A convolution neural network is suitable for complex image characteristics with good robustness. In recent years, it has made a breakthrough in a set of small and efficient neural networks with SqueezeNet, Google Inception and the flattened network method put forward. In order to solve the problems of low automation, low efficiency and high error rate in the traditional remaining oil form recognition algorithm after water flooding in tight oil reservoirs, an image recognition algorithm based on the MobileNets convolutional neural network model was proposed in this paper to achieve accurate recognition of the remaining oil form. Based on traditional image processing methods which, respectively, extracted the whole picture of the different types of remaining oil in the image block, it uses the MobileNets network structure to classify different types of image block and realizes the layered depth convolution neural network system. The experiment result shows that the model can accurately identify the remaining oil forms, and the overall recognition accuracy is up to 83.8% after the convergence of the network model, which infinitely identifies the remaining oil forms in the morphological library, proving the strong generalization and robustness of the model. [ABSTRACT FROM AUTHOR]

Deep learning based on neural networks has been widely used in image recognition, speech recognition, natural language processing, automatic driving, and other fields and has made breakthrough progress. FPGA stands out in the field of accelerated deep learning with its advantages such as flexible architecture and logic units, high energy efficiency ratio, strong compatibility, and low delay. In order to track the latest research results of neural network optimization technology based on FPGA in time and to keep abreast of current research hotspots and application fields, the related technologies and research contents are reviewed. This paper introduces the development history and application fields of some representative neural networks and points out the importance of studying deep learning technology, as well as the reasons and advantages of using FPGA to accelerate deep learning. Several common neural network models are introduced. Moreover, this paper reviews the current mainstream FPGA-based neural network acceleration technology, method, accelerator, and acceleration framework design and the latest research status, pointing out the current FPGA-based neural network application facing difficulties and the corresponding solutions, as well as prospecting the future research directions. We hope that this work can provide insightful research ideas for the researchers engaged in the field of neural network acceleration based on FPGA. [ABSTRACT FROM AUTHOR]

In recent years, deep neural networks have enabled countless innovations in the field of image classification. Encouraged by success in this field, researchers worldwide have demonstrated how to use Convolutional Neural Network techniques in medical imaging problems. In this article, the results were obtained through the use of the EfficientNet in the task of classifying 14 different diseases based on chest X-ray images coming from the NIH (National Institutes of Health) ChestX-ray14 dataset. The approach addresses dataset imbalances by introducing a custom split to ensure fair representation. Binary cross entropy loss is utilized to handle the multi-label difficulty. The model architecture comprises an EfficientNet backbone for feature extraction, succeeded by sequential layers including GlobalAveragePooling, Dense, and BatchNormalization. The main contribution of this paper is a proposed solution that outperforms previous state-of-the-art deep learning models average area under the receiver operating characteristic curve—AUC-ROC (score: 84.28%). The usage of the transfer-learning technique and traditional deep learning engineering techniques was shown to enable us to obtain such results on consumer-class GPUs (graphics processing units). [ABSTRACT FROM AUTHOR]

Recent studies have demonstrated the potential of machine learning methods for fast and accurate mineral classification based on microscope thin sections. Such methods can be extremely useful to support geoscientists during the phases of operational geology, especially when mineralogical and petrological data are fully integrated with other geological and geophysical information. In order to be effective, these methods require robust machine learning models trained on pre-labeled data. Furthermore, it is mandatory to optimize the hyper-parameters of the machine learning techniques in order to guarantee optimal classification accuracy and reliability. Nowadays, deep learning algorithms are widely applied for image analysis and automatic classification in a large range of Earth disciplines, including mineralogy, petrography, paleontology, well-log analysis, geophysical imaging, and so forth. The main reason for the recognized effectiveness of deep learning algorithms for image analysis is that they are able to quickly learn complex representations of images and patterns within them. Differently from traditional image-processing techniques based on handcrafted features, deep learning models automatically learn and extract features from the data, capturing, in almost real-time, complex relationships and patterns that are difficult to manually define. Many different types of deep learning models can be used for image analysis and classification, including fully connected deep neural networks (FCNNs), convolutional neural networks (CNNs or ConvNet), and residual networks (ResNets). In this paper, we compare some of these techniques and verify their effectiveness on the same dataset of mineralogical thin sections. We show that the different deep learning methods are all effective techniques in recognizing and classifying mineral images directly in the field, with ResNets outperforming the other techniques in terms of accuracy and precision. In addition, we compare the performance of deep learning techniques with different machine learning algorithms, including random forest, naive Bayes, adaptive boosting, support vector machine, and decision tree. Using quantitative performance indexes as well as confusion matrixes, we demonstrate that deep neural networks show generally better classification performances than the other approaches. Furthermore, we briefly discuss how to expand the same workflow to other types of images and geo-data, showing how this deep learning approach can be generalized to a multiscale/multipurpose methodology addressed to the analysis and automatic classification of multidisciplinary information. This article has tutorial purposes, too. For that reason, we will explain, with a didactical level of detail, all the key steps of the workflow. [ABSTRACT FROM AUTHOR]

Deep neural networks have achieved remarkable success in various fields of artificial intelligence. However, these models, which contain a large number of parameters, are widely distributed and disseminated by researchers, engineers, and even unauthorized users. Except for intelligent tasks, typically overparameterized deep neural networks have become new digital covers for data hiding, which may pose significant security challenges to AI systems. To address this issue, this paper proposes a symmetric steganalysis scheme specifically designed for neural networks trained for image classification tasks. The proposed method focuses on detecting the presence of additional data without access to the internal structure or parameters of the host network. It employs a well-designed method based on histogram distribution to find the optimal decision threshold, with a symmetric determining rule where the original networks and stego networks undergo two highly symmetrical flows to generate the classification labels; the method has been shown to be practical and effective. SVM and ensemble classifiers were chosen as the binary classifier for their applicability to feature vectors output from neural networks based on different datasets and network structures. This scheme is the first of its kind, focusing on steganalysis for neural networks based on the distribution of network output, compared to conventional digital media such as images, audio, and video. Overall, the proposed scheme offers a promising approach to enhancing the security of deep neural networks against data hiding attacks. [ABSTRACT FROM AUTHOR]

A bite from a bug may expose the affected person to serious, life-threatening conditions, which may require immediate medical attention. The identification of the bug bite may be challenging even for experienced medical personnel due to the different manifestations of the bites and similarity to other skin conditions. This motivated our work on a computer-aided system that offers information on the bug bite based on the classification of bug bite images. Recently, there have been significant advances of methods for image classification for the detection of various skin conditions. However, there are very few sources that discuss the classification of bug bites. The goal of our research is to fill in this gap in the literature and offer a comprehensive approach for the analysis of this topic. This includes (1) the creation of a dataset that is larger than those considered in the related sources; (2) the exploration and analysis of the application of pre-trained state-of-the-art deep learning architectures with transfer learning, used in this study to overcome the challenges of low-size datasets and computational burden; (3) the further improvement of the classification performance of the individual CNNs by proposing an ensemble of models, and finally, (4) the implementation and description of an end-to-end system for bug bite classification from images taken with mobile phones, which should be beneficial to the medical personnel in the diagnostic process. In this paper, we give a detailed discussion of the models' architecture, back-end architecture, and performance. According to the general evaluation metrics, DenseNet169 with an accuracy of 78% outperformed the other individual CNN models. However, the overall best performance (accuracy of 86%) was achieved by the proposed stacking ensemble model. These results are better than the results in the limited related work. Additionally, they show that deep CNNs and transfer learning can be successfully applied to the problem of the classification of bug bites. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNNs), the integration of neural networks (NNs) and deep learning (DL), have proven highly efficient in executing numerous complex tasks, such as data and image classification. Because the multilayer in a nonlinearly separable data structure is not transparent, it is critical to develop a specific data classification model from a new and unexpected dataset. In this paper, we propose a novel approach using the concepts of DNN and decision tree (DT) for classifying nonlinear data. We first developed a decision tree-based neural network (DTBNN) model. Next, we extend our model to a decision tree-based deep neural network (DTBDNN), in which the multiple hidden layers in DNN are utilized. Using DNN, the DTBDNN model achieved higher accuracy compared to the related and relevant approaches. Our proposal achieves the optimal trainable weights and bias to build an efficient model for nonlinear data classification by combining the benefits of DT and NN. By conducting in-depth performance evaluations, we demonstrate the effectiveness and feasibility of the proposal by achieving good accuracy over different datasets. [ABSTRACT FROM AUTHOR]

In the past decade, deep neural networks, particularly convolutional neural networks, have revolutionised computer vision. However, all deep learning models may require a large amount of data so as to achieve satisfying results. Unfortunately, the availability of sufficient amounts of data for real-world problems is not always possible, and it is well recognised that a paucity of data easily results in overfitting. This issue may be addressed through several approaches, one of which is data augmentation. In this paper, we survey the existing data augmentation techniques in computer vision tasks, including segmentation and classification, and suggest new strategies. In particular, we introduce a way of implementing data augmentation by using local information in images. We propose a parameter-free and easy to implement strategy, the random local rotation strategy, which involves randomly selecting the location and size of circular regions in the image and rotating them with random angles. It can be used as an alternative to the traditional rotation strategy, which generally suffers from irregular image boundaries. It can also complement other techniques in data augmentation. Extensive experimental results and comparisons demonstrated that the new strategy consistently outperformed its traditional counterparts in, for example, image classification. [ABSTRACT FROM AUTHOR]

Featured Application: This work will be materialized into the Bioimage Diagnostic system in Kaohsiung Branch of Chang Gung Hospital, Taiwan. Image recognition through the use of deep learning (DL) techniques has recently become a hot topic in many fields. Especially for bioimage informatics, DL-based image recognition has been successfully used in several applications, such as cancer and fracture detection. However, few previous studies have focused on detecting scaphoid fractures, and the related effectiveness is also not significant. Aimed at this issue, in this paper, we present a two-stage method for scaphoid fracture recognition by conducting an effectiveness analysis of numerous state-of-the-art artificial neural networks. In the first stage, the scaphoid bone is extracted from the radiograph using object detection techniques. Based on the object extracted, several convolutional neural networks (CNNs), with or without transfer learning, are utilized to recognize the segmented object. Finally, the analytical details on a real data set are given, in terms of various evaluation metrics, including sensitivity, specificity, precision, F1-score, area under the receiver operating curve (AUC), kappa, and accuracy. The experimental results reveal that the CNNs with transfer learning are more effective than those without transfer learning. Moreover, DenseNet201 and ResNet101 are found to be more promising than the other methods, on average. According to the experimental results, DenseNet201 and ResNet101 can be recommended as considerable solutions for scaphoid fracture detection within a bioimage diagnostic system. [ABSTRACT FROM AUTHOR]

Ship fire is one of the greatest dangers to ship navigation safety. Nevertheless, typical detection methods have limited detection effectiveness and accuracy due to distance restrictions and ship motion. Although the issue can be addressed by image recognition algorithms based on deep learning, the computational complexity and efficiency for ship detection are tough. This paper proposes a lightweight target identification technique based on the modified YOLOv4-tiny algorithm for the precise and efficient detection of ship fires, taking into account the distinctive characteristics of ship fires and the marine environment. Initially, a multi-scale detection technique is applied to broaden the detection range and integrate deep semantic information, thereby enhancing the feature information of small targets and obscured objects and improving the detection precision. Then, the proposed algorithm employs the SE attention mechanism for inter-channel feature fusion to improve the capability of feature extraction and the precision of ship fire detection. Last but not least, picture transformation and migration learning are added to the small ship fire dataset to accelerate the convergence pace, improve the convergence effect, and reduce dataset dependence. The simulation experiments reveal that the proposed I-YOLOv4-tiny + SE model outperforms the benchmark algorithm in terms of ship fire detection accuracy and detection efficiency and that it satisfies the real-time ship fire warning criteria in demanding maritime environments. [ABSTRACT FROM AUTHOR]

The article discusses the advancements and challenges in the field of image recognition, particularly in the context of deep learning and neural networks. It highlights the significant progress made in object detection and image classification tasks, surpassing human capabilities. The article also explores various topics within image recognition, such as point cloud deep learning, pixelwise semantic segmentation, zeroshot learning, model optimization, and multimodal applications. It emphasizes the need for further research and innovation in these areas to overcome existing challenges and improve the effectiveness and versatility of deep visual systems. The article concludes by expressing gratitude to the authors and reviewers for their contributions to the field. [Extracted from the article]

Environmental Sound Recognition has become a relevant application for smart cities. Such an application, however, demands the use of trained machine learning classifiers in order to categorize a limited set of audio categories. Although classical machine learning solutions have been proposed in the past, most of the latest solutions that have been proposed toward automated and accurate sound classification are based on a deep learning approach. Deep learning models tend to be large, which can be problematic when considering that sound classifiers often have to be embedded in resource constrained devices. In this paper, a classical machine learning based classifier called MosAIc, and a lighter Convolutional Neural Network model for environmental sound recognition, are proposed to directly compete in terms of accuracy with the latest deep learning solutions. Both approaches are evaluated in an embedded system in order to identify the key parameters when placing such applications on constrained devices. The experimental results show that classical machine learning classifiers can be combined to achieve similar results to deep learning models, and even outperform them in accuracy. The cost, however, is a larger classification time. [ABSTRACT FROM AUTHOR]

Creating a widely excepted model on the measure of intelligence became inevitable due to the existence of an abundance of different intelligent systems. Measuring intelligence would provide feedback for the developers and ultimately lead us to create better artificial systems. In the present paper, we show a solution where learning as a process is examined, aiming to detect pre-written solutions and separate them from the knowledge acquired by the system. In our approach, we examine image recognition software by executing different transformations on objects and detect if the software was resilient to it. A system with the required intelligence is supposed to become resilient to the transformation after experiencing it several times. The method is successfully tested on a simple neural network, which is not able to learn most of the transformations examined. The method can be applied to any image recognition software to test its abstraction capabilities. [ABSTRACT FROM AUTHOR]

First-order algorithms have long dominated the training of deep neural networks, excelling in tasks like image classification and natural language processing. Now there is a compelling opportunity to explore alternatives that could outperform current state-of-the-art results. From the estimation theory, the Extended Kalman Filter (EKF) arose as a viable alternative and has shown advantages over backpropagation methods. Current computational advances offer the opportunity to review algorithms derived from the EKF, almost excluded from the training of convolutional neural networks. This article revisits an approach of the EKF with decoupling and it brings the Fully Decoupled Extended Kalman Filter (FDEKF) for training convolutional neural networks in image classification tasks. The FDEKF is a second-order algorithm with some advantages over the first-order algorithms, so it can lead to faster convergence and higher accuracy, due to a higher probability of finding the global optimum. In this research, experiments are conducted on well-known datasets that include Fashion, Sports, and Handwritten Digits images. The FDEKF shows faster convergence compared to other algorithms such as the popular Adam optimizer, the sKAdam algorithm, and the reduced extended Kalman filter. Finally, motivated by the finding of the highest accuracy of FDEKF with images of natural scenes, we show its effectiveness in another experiment focused on outdoor terrain recognition. [ABSTRACT FROM AUTHOR]

The complexity of deep neural network models (DNNs) severely limits their application on devices with limited computing and storage resources. Knowledge distillation (KD) is an attractive model compression technology that can effectively alleviate this problem. Multi-teacher knowledge distillation (MKD) aims to leverage the valuable and diverse knowledge distilled by multiple teacher networks to improve the performance of the student network. Existing approaches typically rely on simple methods such as averaging the prediction logits or using sub-optimal weighting strategies to fuse distilled knowledge from multiple teachers. However, employing these techniques cannot fully reflect the importance of teachers and may even mislead student's learning. To address this issue, we propose a novel Decoupled Multi-Teacher Knowledge Distillation based on Entropy (DE-MKD). DE-MKD decouples the vanilla knowledge distillation loss and assigns adaptive weights to each teacher to reflect its importance based on the entropy of their predictions. Furthermore, we extend the proposed approach to distill the intermediate features from multiple powerful but cumbersome teachers to improve the performance of the lightweight student network. Extensive experiments on the publicly available CIFAR-100 image classification benchmark dataset with various teacher-student network pairs demonstrated the effectiveness and flexibility of our approach. For instance, the VGG8|ShuffleNetV2 model trained by DE-MKD reached 75.25%|78.86% top-one accuracy when choosing VGG13|WRN40-2 as the teacher, setting new performance records. In addition, surprisingly, the distilled student model outperformed the teacher in both teacher-student network pairs. [ABSTRACT FROM AUTHOR]

Despite the considerable advancements in automated identification methods of highway hidden distress with ground-penetrating radar (GPR) images, there still exist challenges in realizing automated identification of highway hidden distress owing to the quantity, variability, and reliability of the distress samples and diversity of classification models. Firstly, the dataset collected contains 31,640 samples categorized into four categories: interlayer debonding, interlayer loosening, interlayer water seepage, and structural loosening from 1500 km highway, for obtaining larger enough samples and covering the variable range of distress samples. Secondly, the distresses were labeled by experienced experts, and the labels were verified with drilled cores to ensure their reliability. Lastly, 18 exemplary convolutional neural network (CNN) models from 8 different architectures were evaluated using evaluation metrics such as precision, recall, and f1-score. Further, confusion matrix and Grad-CAM techniques were utilized to analyze these models. The experimental results show that VGG13 performed most prominently and stably, while the lightweight network SqueezeNet1_1 performed particularly well with a batch size of 64. Furthermore, this study indicates that models with fewer layers can achieve comparable or better performance than deeper models. [ABSTRACT FROM AUTHOR]

With the development of deep learning in the field of computer vision, convolutional neural network models and attention mechanisms have been widely applied in SAR image target recognition. The improvement of convolutional neural network attention in existing SAR image target recognition focuses on spatial and channel information but lacks research on the relationship and recognition mechanism between spatial and channel information. In response to this issue, this article proposes a hybrid attention module and introduces a Mixed Attention (MA) mechanism module in the MobileNetV2 network. The proposed MA mechanism fully considers the comprehensive calculation of spatial attention (SPA), channel attention (CHA), and coordinated attention (CA). It can input feature maps for comprehensive weighting to enhance the features of the regions of interest, in order to improve the recognition rate of vehicle targets in SAR images.The superiority of our algorithm was verified through experiments on the MSTAR dataset. [ABSTRACT FROM AUTHOR]

In the current field of disease risk prediction research, there are many methods of using servers for centralized computing to train and infer prediction models. However, this centralized computing method increases storage space, the load on network bandwidth, and the computing pressure on the central server. In this article, we design an image preprocessing method and propose a lightweight neural network model called Linge (Lightweight Neural Network Models for the Edge). We propose a distributed intelligent edge computing technology based on the federated learning algorithm for disease risk prediction. The intelligent edge computing method we proposed for disease risk prediction directly performs prediction model training and inference at the edge without increasing storage space. It also reduces the load on network bandwidth and reduces the computing pressure on the server. The lightweight neural network model we designed has only 7.63 MB of parameters and only takes up 155.28 MB of memory. In the experiment with the Linge model compared with the EfficientNetV2 model, the accuracy and precision increased by 2%, the recall rate increased by 1%, the specificity increased by 4%, the F1 score increased by 3%, and the AUC (Area Under the Curve) value increased by 2%. [ABSTRACT FROM AUTHOR]

Susceptibility to domain changes for image classification hinders the application and development of deep neural networks. Domain adaptation (DA) makes use of domain-invariant characteristics to improve the performance of a model trained on labeled data from one domain (source domain) on an unlabeled domain (target) with a different data distribution. But existing DA methods simply use pretrained models (e.g., AlexNet, ResNet) for feature extraction, which are convolutional models that are trapped in localized features and fail to acquire long-distance dependencies. Furthermore, many approaches depend too much on pseudo-labels, which can impair adaptation efficiency and lead to unstable and inconsistent results. In this research, we present S2AC, a novel approach for unsupervised deep domain adaptation, that makes use of a stacked attention architecture as a feature map extractor. Our method can fuse domain discrepancy with minimizing a linear transformation of the second statistics (covariances) extended by the p-norm, while simultaneously designing pretext tasks on heuristics to improve the generality of the learning representation. In addition, we have developed a new trainable relative position embedding that not only reduces the model parameters but also enhances model accuracy and expedites the training process. To illustrate our method's efficacy and controllability, we designed extensive experiments based on the Office31, Office_Caltech_10, and OfficeHome datasets. To the best of our knowledge, the proposed method is the first attempt at incorporating attention-based networks and self-supervised learning for image domain adaptation, and has shown promising results. [ABSTRACT FROM AUTHOR]

With the development of few-shot learning, significant progress has been achieved in hyperspectral image classification using related networks, leading to improved classification outcomes. However, practical few-shot hyperspectral image classification encounters challenges such as network overfitting and insufficient feature extraction during the model training process. To address these issues, we propose a model called CRSSNet (Convolutional Residuals and SAM Siamese Networks) for few-shot hyperspectral image classification. In this model, we deepen the network depth and employ the convolutional residual technique to enhance the feature extraction capabilities and alleviate the problem of network gradient degradation. Additionally, we introduce the Spatial Attention Mechanism (SAM) to effectively leverage spatial information features in hyperspectral images. Lastly, metric learning is employed by comparing the distance between two output feature vectors to determine the label category. Experimental results demonstrate that our method achieves superior classification performance compared to other methods. [ABSTRACT FROM AUTHOR]

Neural network models play an important role in crop extraction based on remote sensing data. However, when dealing with high-dimensional remote sensing data, these models are susceptible to performance degradation. In order to address the challenges associated with multi-source Gaofen satellite data, a novel method is proposed for dimension reduction and crop classification. This method combines the benefits of the stacked autoencoder network for data dimensionality reduction, and the convolutional neural network for classification. By leveraging the advantages of multi-dimensional remote sensing information, and mitigating the impact of dimensionality on the classification accuracy, this method aims to improve the effectiveness of crop classification. The proposed method was applied to the extraction of crop-planting areas in the Yangling Agricultural Demonstration Zone, using multi-temporal spectral data collected from the Gaofen satellites. The results demonstrate that the fusion network, which extracts low-dimensional characteristics, offers advantages in classification accuracy. At the same time, the proposed model is compared with methods such as the decision tree (DT), random forest (RF), support vector machine (SVM), hyperspectral image classification based on a convolutional neural network (HICCNN), and a characteristic selection classification method based on a convolutional neural network (CSCNN). The overall accuracy of the proposed method can reach 98.57%, which is 7.95%, 4.69%, 5.68%, 1.21%, and 1.10% higher than the above methods, respectively. The effectiveness of the proposed model was verified through experiments. Additionally, the model demonstrates a strong robustness when classifying based on new data. When extracting the crop area of the entire Yangling District, the errors for wheat and corn are only 9.6% and 6.3%, respectively, and the extraction results accurately reflect the actual planting situation of crops. [ABSTRACT FROM AUTHOR]

Currently, deep learning networks, with particular regard to convolutional neural network models, are typically exploited for biomedical image classification. One of the disadvantages of deep learning is that is extremely expensive to train due to complex data models. Extreme learning machine has recently emerged which, as shown in experimental studies, can produce an acceptable predictive performance in several classification tasks, and at a much lower training cost compared to deep learning networks that are trained by backpropagation. We propose a method devoted to exploring the possibility of considering extreme learning machines for biomedical classification tasks. Binary and multiclass classification in four case studies are considered to demonstrate the effectiveness of extreme learning machine, considering the biomedical images acquired with the dermatoscope and with the blood cell microscope, showing that the extreme learning machine can be successfully applied for biomedical image classification. [ABSTRACT FROM AUTHOR]

Deep learning models have been used in creating various effective image classification applications. However, they are vulnerable to adversarial attacks that seek to misguide the models into predicting incorrect classes. Our study of major adversarial attack models shows that they all specifically target and exploit the neural networking structures in their designs. This understanding led us to develop a hypothesis that most classical machine learning models, such as random forest (RF), are immune to adversarial attack models because they do not rely on neural network design at all. Our experimental study of classical machine learning models against popular adversarial attacks supports this hypothesis. Based on this hypothesis, we propose a new adversarial-aware deep learning system by using a classical machine learning model as the secondary verification system to complement the primary deep learning model in image classification. Although the secondary classical machine learning model has less accurate output, it is only used for verification purposes, which does not impact the output accuracy of the primary deep learning model, and, at the same time, can effectively detect an adversarial attack when a clear mismatch occurs. Our experiments based on the CIFAR-100 dataset show that our proposed approach outperforms current state-of-the-art adversarial defense systems. [ABSTRACT FROM AUTHOR]

Automatic classification of histopathological images plays an important role in computer-aided diagnosis systems. The automatic classification model of histopathological images based on deep neural networks has received widespread attention. However, the performance of deep models is affected by many factors, such as training hyperparameters, model structure, dataset quality, and training cost. In order to reduce the impact of the above factors on model training and reduce the training and inference costs of the model, we propose a novel method based on model fusion in the weight space, which is inspired by stochastic weight averaging and model soup. We use the cyclical learning rate (CLR) strategy to fine-tune the ingredient models and propose a ranking strategy based on accuracy and diversity for candidate model selection. Compared to the single model, the weight fusion of ingredient models can obtain a model whose performance is closer to the expected value of the error basin, which may improve the generalization ability of the model. Compared to the ensemble model with n base models, the testing cost of the proposed model is theoretically 1 / n of that of the ensemble model. Experimental results on two histopathological image datasets show the effectiveness of the proposed model in comparison to baseline ones, including ResNet, VGG, DenseNet, and their ensemble versions. [ABSTRACT FROM AUTHOR]

The automatic character recognition of historic documents gained more attention from scholars recently, due to the big improvements in computer vision, image processing, and digitization. While Neural Networks, the current state-of-the-art models used for image recognition, are very performant, they typically suffer from using large amounts of training data. In our study we manually built our own relatively small dataset of 404 characters by cropping letter images from a popular historic manuscript, the Electronic Beowulf. To compensate for the small dataset we use ImageDataGenerator, a Python library was used to augment our Beowulf manuscript's dataset. The training dataset was augmented once, twice, and thrice, which we call resampling 1, resampling 2, and resampling 3, respectively. To classify the manuscript's character images efficiently, we developed a customized Convolutional Neural Network (CNN) model. We conducted a comparative analysis of the results achieved by our proposed model with other machine learning (ML) models such as support vector machine (SVM), K-nearest neighbor (KNN), decision tree (DT), random forest (RF), and XGBoost. We used pretrained models such as VGG16, MobileNet, and ResNet50 to extract features from character images. We then trained and tested the above ML models and recorded the results. Moreover, we validated our proposed CNN model against the well-established MNIST dataset. Our proposed CNN model achieves very good recognition accuracies of 88.67%, 90.91%, and 98.86% in the cases of resampling 1, resampling 2, and resampling 3, respectively, for the Beowulf manuscript's data. Additionally, our CNN model achieves the benchmark recognition accuracy of 99.03% for the MNIST dataset. [ABSTRACT FROM AUTHOR]

A framework combining two powerful tools of hyperspectral imaging and deep learning for the processing and classification of hyperspectral images (HSI) of rice seeds is presented. A seed-based approach that trains a three-dimensional convolutional neural network (3D-CNN) using the full seed spectral hypercube for classifying the seed images from high day and high night temperatures, both including a control group, is developed. A pixel-based seed classification approach is implemented using a deep neural network (DNN). The seed and pixel-based deep learning architectures are validated and tested using hyperspectral images from five different rice seed treatments with six different high temperature exposure durations during day, night, and both day and night. A stand-alone application with Graphical User Interfaces (GUI) for calibrating, preprocessing, and classification of hyperspectral rice seed images is presented. The software application can be used for training two deep learning architectures for the classification of any type of hyperspectral seed images. The average overall classification accuracy of 91.33% and 89.50% is obtained for seed-based classification using 3D-CNN for five different treatments at each exposure duration and six different high temperature exposure durations for each treatment, respectively. The DNN gives an average accuracy of 94.83% and 91% for five different treatments at each exposure duration and six different high temperature exposure durations for each treatment, respectively. The accuracies obtained are higher than those presented in the literature for hyperspectral rice seed image classification. The HSI analysis presented here is on the Kitaake cultivar, which can be extended to study the temperature tolerance of other rice cultivars. [ABSTRACT FROM AUTHOR]

Deep learning requires a large amount of datasets to train deep neural network models for specific tasks, and thus training of a new model is a very costly task. Research on transfer networks used to reduce training costs will be the next turning point in deep learning research. The use of source task models to help reduce the training costs of the target task models, especially heterogeneous systems, is a problem we are studying. In order to quickly obtain an excellent target task model driven by the source task model, we propose a novel transfer learning approach. The model linearly transforms the feature mapping of the target domain and increases the weight value for feature matching to realize the knowledge transfer between heterogeneous networks and add a domain discriminator based on the principle of generative adversarial to speed up feature mapping and learning. Most importantly, this paper proposes a new objective function optimization scheme to complete the model training. It successfully combines the generative adversarial network with the weight feature matching method to ensure that the target model learns the most beneficial features from the source domain for its task. Compared with the previous transfer algorithm, our training results are excellent under the same benchmark for image recognition tasks. [ABSTRACT FROM AUTHOR]

Deep learning is a sub-branch of artificial intelligence that acquires knowledge by training a neural network. It has many applications in the field of banking, automobile industry, agriculture, and healthcare industry. Deep learning has played a significant role in solving complex tasks related to computer vision, such as image classification, natural language processing, and object detection. On the other hand, optimizers also play an intrinsic role in training the deep learning model. Recent studies have proposed many deep learning models, such as VGG, ResNet, DenseNet, and ImageNet. In addition, there are many optimizers such as stochastic gradient descent (SGD), Adam, AdaDelta, Adabelief, and AdaMax. In this study, we have selected those models that require lower hardware requirements and shorter training times, which facilitates the overall training process. We have modified the Adam based optimizers and minimized the cyclic path. We have removed an additional hyper-parameter from RMSProp and observed that the optimizer works with various models. The learning rate is set to minimum and constant. The initial weights are updated after each epoch, which helps to improve the accuracy of the model. We also changed the position of the epsilon in the default Adam optimizer. By changing the position of the epsilon, it accumulates the updating process. We used various models with SGD, Adam, RMSProp, and the proposed optimization technique. The results indicate that the proposed method is effective in achieving the accuracy and works well with the state-of-the-art architectures. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNNs) can improve the image analysis and interpretation of remote sensing technology by extracting valuable information from images, and has extensive applications such as military affairs, agriculture, environment, transportation, and urban division. The DNNs for object detection can identify and analyze objects in remote sensing images through fruitful features of images, which improves the efficiency of image processing and enables the recognition of large-scale remote sensing images. However, many studies have shown that deep neural networks are vulnerable to adversarial attack. After adding small perturbations, the generated adversarial examples will cause deep neural network to output undesired results, which will threaten the normal recognition and detection of remote sensing systems. According to the application scenarios, attacks can be divided into the digital domain and the physical domain, the digital domain attack is directly modified on the original image, which is mainly used to simulate the attack effect, while the physical domain attack adds perturbation to the actual objects and captures them with device, which is closer to the real situation. Attacks in the physical domain are more threatening, however, existing attack methods generally generate the patch with bright style and a large attack range, which is easy to be observed by human vision. Our goal is to generate a natural patch with a small perturbation area, which can help some remote sensing images used in the military to avoid detection by object detectors and im-perceptible to human eyes. To address the above issues, we generate a rust-style adversarial patch generation framework based on style transfer. The framework takes a heat map-based interpretability method to obtain key areas of target recognition and generate irregular-shaped natural-looking patches to reduce the disturbance area and alleviates suspicion from humans. To make the generated adversarial examples have a higher attack success rate in the physical domain, we further improve the robustness of the adversarial patch through data augmentation methods such as rotation, scaling, and brightness, and finally, make it impossible for the object detector to detect the camouflage patch. We have attacked the YOLOV3 detection network on multiple datasets. The experimental results show that our model has achieved a success rate of 95.7% in the digital domain. We also conduct physical attacks in indoor and outdoor environments and achieve an attack success rate of 70.6% and 65.3%, respectively. The structural similarity index metric shows that the adversarial patches generated are more natural than existing methods. [ABSTRACT FROM AUTHOR]

Nearshore sandbars characterize many sandy coasts, and unravelling their dynamics is crucial to understanding nearshore sediment pathways. Sandbar morphologies exhibit complex patterns that can be classified into distinct states. The tremendous progress in data-driven learning in image recognition has recently led to the first automated classification of single-barred beach states from Argus imagery using a Convolutional Neural Network (CNN). Herein, we extend this method for the classification of beach states in a double-barred system. We used transfer learning to fine-tune the pre-trained network of ResNet50. Our data consisted of labelled single-bar time-averaged images from the beaches of Narrabeen (Australia) and Duck (US), complemented by 9+ years of daily averaged low-tide images of the double-barred beach of the Gold Coast (Australia). We assessed seven different CNNs, of which each model was tested on the test data from the location where its training data came from, the self-tests, and on the test data of alternate, unseen locations, the transfer-tests. When the model trained on the single-barred data of both Duck and Narrabeen was tested on unseen data of the double-barred Gold Coast, we achieved relatively low performances as measured by F1 scores. In contrast, models trained with only the double-barred beach data showed comparable skill in the self-tests with that of the single-barred models. We incrementally added data with labels from the inner or outer bar of the Gold Coast to the training data from both single-barred beaches, and trained models with both single- and double-barred data. The tests with these models showed that which bar the labels used for training the model mattered. The training with the outer bar labels led to overall higher performances, except at the inner bar. Furthermore, only 10% of additional data with the outer bar labels was needed for reasonable transferability, compared to the 20% of additional data needed with the inner bar labels. Additionally, when trained with data from multiple locations, more data from a new location did not always positively affect the model's performance on other locations. However, the larger diversity of images coming from more locations allowed the transferability of the model to the locations from where new training data were added. [ABSTRACT FROM AUTHOR]

The abnormal behavior of cockpit pilots during the manipulation process is an important incentive for flight safety, but the complex cockpit environment limits the detection accuracy, with problems such as false detection, missed detection, and insufficient feature extraction capability. This article proposes a method of abnormal pilot driving behavior detection based on the improved YOLOv4 deep learning algorithm and by integrating an attention mechanism. Firstly, the semantic image features are extracted by running the deep neural network structure to complete the image and video recognition of pilot driving behavior. Secondly, the CBAM attention mechanism is introduced into the neural network to solve the problem of gradient disappearance during training. The CBAM mechanism includes both channel and spatial attention processes, meaning the feature extraction capability of the network can be improved. Finally, the features are extracted through the convolutional neural network to monitor the abnormal driving behavior of pilots and for example verification. The conclusion shows that the deep learning algorithm based on the improved YOLOv4 method is practical and feasible for the monitoring of the abnormal driving behavior of pilots during the flight maneuvering phase. The experimental results show that the improved YOLOv4 recognition rate is significantly higher than the unimproved algorithm, and the calling phase has a mAP of 87.35%, an accuracy of 75.76%, and a recall of 87.36%. The smoking phase has a mAP of 87.35%, an accuracy of 85.54%, and a recall of 85.54%. The conclusion shows that the deep learning algorithm based on the improved YOLOv4 method is practical and feasible for the monitoring of the abnormal driving behavior of pilots in the flight maneuvering phase. This method can quickly and accurately identify the abnormal behavior of pilots, providing an important theoretical reference for abnormal behavior detection and risk management. [ABSTRACT FROM AUTHOR]