Search

Image watermarking often involves the use of handheld devices under non-structured conditions for authentication purposes, particularly in the print-cam process where smartphone cameras are used to capture watermarked printed images. However, these images frequently suffer from perspective distortions, making them unsuitable for automated information detection. To address this issue, Cam-Unet, an end-to-end neural network architecture, is presented to predict the mapping from distorted images to rectified ones, specifically tailored for print-cam challenges applied to ID images. Given the limited availability of large-scale real datasets containing ground truth distortions, we created an extensive synthetic dataset by subjecting undistorted images to print-cam attacks. The proposed network is trained on this dataset, using various data augmentation techniques to improve its generalization capabilities. Accordingly, this paper presents an image watermarking system for the print-cam process. The approach combines Fourier transform-based watermarking with Cam-Unet as perspective distortion correction. Results show that the proposed method outperforms existing watermarking approaches typically employed to counter print-cam attacks and achieves an optimal balance between efficiency and cost-effectiveness. [ABSTRACT FROM AUTHOR]

Human–object interaction (HOI) detection identifies a "set of interactions" in an image involving the recognition of interacting instances and the classification of interaction categories. The complexity and variety of image content make this task challenging. Recently, the Transformer has been applied in computer vision and received attention in the HOI detection task. Therefore, this paper proposes a novel Part Refinement Tandem Transformer (PRTT) for HOI detection. Unlike the previous Transformer-based HOI method, PRTT utilizes multiple decoders to split and process rich elements of HOI prediction and introduces a new part state feature extraction (PSFE) module to help improve the final interaction category classification. We adopt a novel prior feature integrated cross-attention (PFIC) to utilize the fine-grained partial state semantic and appearance feature output obtained by the PSFE module to guide queries. We validate our method on two public datasets, V-COCO and HICO-DET. Compared to state-of-the-art models, the performance of detecting human–object interaction is significantly improved by the PRTT. [ABSTRACT FROM AUTHOR]

This paper proposes a new method for recognizing, extracting, and processing Phase-Resolved Partial Discharge (PRPD) patterns from two-dimensional plots to identify specific defect types affecting electrical equipment without human intervention while retaining the principals that make PRPD analysis an effective diagnostic technique. The proposed method does not rely on training complex deep learning algorithms which demand substantial computational resources and extensive datasets that can pose significant hurdles for the application of on-line partial discharge monitoring. Instead, the developed Cosine Cluster Net (CCNet) model, which is an image processing pipeline, can extract and process patterns from any two-dimensional PRPD plot before employing the cosine similarity function to measure the likeness of the patterns to predefined templates of known defect types. The PRPD pattern recognition capabilities of the model were tested using several manually classified PRPD images available in the existing literature. The model consistently produced similarity scores that identified the same defect type as the one from the manual classification. The successful defect type reporting from the initial trials of the CCNet model together with the speed of the identification, which typically does not exceed four seconds, indicates potential for real-time applications. [ABSTRACT FROM AUTHOR]

In recent years, deep learning methods have achieved remarkable success in hyperspectral image classification (HSIC), and the utilization of convolutional neural networks (CNNs) has proven to be highly effective. However, there are still several critical issues that need to be addressed in the HSIC task, such as the lack of labeled training samples, which constrains the classification accuracy and generalization ability of CNNs. To address this problem, a deep multi-scale attention fusion network (DMAF-NET) is proposed in this paper. This network is based on multi-scale features and fully exploits the deep features of samples from multiple levels and different perspectives with an aim to enhance HSIC results using limited samples. The innovation of this article is mainly reflected in three aspects: Firstly, a novel baseline network for multi-scale feature extraction is designed with a pyramid structure and densely connected 3D octave convolutional network enabling the extraction of deep-level information from features at different granularities. Secondly, a multi-scale spatial–spectral attention module and a pyramidal multi-scale channel attention module are designed, respectively. This allows modeling of the comprehensive dependencies of coordinates and directions, local and global, in four dimensions. Finally, a multi-attention fusion module is designed to effectively combine feature mappings extracted from multiple branches. Extensive experiments on four popular datasets demonstrate that the proposed method can achieve high classification accuracy even with fewer labeled samples. [ABSTRACT FROM AUTHOR]

To develop socially assistive robots for monitoring older adults at home, a sensor is required to identify residents and capture activities within the room without violating privacy. We focused on 2D Light Detection and Ranging (2D-LIDAR) capable of robustly measuring human contours in a room. While horizontal 2D contour data can provide human location, identifying humans and activities from these contours is challenging. To address this issue, we developed novel methods using deep learning techniques. This paper proposes methods for person identification and activity estimation in a room using contour point clouds captured by a single 2D-LIDAR at hip height. In this approach, human contours were extracted from 2D-LIDAR data using density-based spatial clustering of applications with noise. Subsequently, the person and activity within a 10-s interval were estimated employing deep learning techniques. Two deep learning models, namely Long Short-Term Memory (LSTM) and image classification (VGG16), were compared. In the experiment, a total of 120 min of walking data and 100 min of additional activities (door opening, sitting, and standing) were collected from four participants. The LSTM-based and VGG16-based methods achieved accuracies of 65.3% and 89.7%, respectively, for person identification among the four individuals. Furthermore, these methods demonstrated accuracies of 94.2% and 97.9%, respectively, for the estimation of the four activities. Despite the 2D-LIDAR point clouds at hip height containing small features related to gait, the results indicate that the VGG16-based method has the capability to identify individuals and accurately estimate their activities. [ABSTRACT FROM AUTHOR]

It is difficult for traditional signal-recognition methods to effectively classify and identify multiple emitter signals in a low SNR environment. This paper proposes a multi-emitter signal-feature-sorting and recognition method based on low-order cyclic statistics CWD time-frequency images and the YOLOv5 deep network model, which can quickly dissociate, label, and sort the multi-emitter signal features in the time-frequency domain under a low SNR environment. First, the denoised signal is extracted based on the low-order cyclic statistics of the typical modulation types of radiation source signals. Second, the time-frequency graph of multisource signals was obtained through CWD time-frequency analysis. The cyclic frequency was controlled to balance the noise suppression effect and operation time to achieve noise suppression of multisource signals at a low SNR. Finally, the YOLOv5s deep network model is used as a classifier to sort and identify the received signals from multiple radiation sources. The method proposed in this paper has high real-time performance. It can identify the radiation source signals of different modulation types with high accuracy under the condition of a low SNR. [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]

Detecting and classifying vehicles as objects from images and videos is challenging in appearance-based representation, yet plays a significant role in the substantial real-time applications of Intelligent Transportation Systems (ITSs). The rapid development of Deep Learning (DL) has resulted in the computer-vision community demanding efficient, robust, and outstanding services to be built in various fields. This paper covers a wide range of vehicle detection and classification approaches and the application of these in estimating traffic density, real-time targets, toll management and other areas using DL architectures. Moreover, the paper also presents a detailed analysis of DL techniques, benchmark datasets, and preliminaries. A survey of some vital detection and classification applications, namely, vehicle detection and classification and performance, is conducted, with a detailed investigation of the challenges faced. The paper also addresses the promising technological advancements of the last few years. [ABSTRACT FROM AUTHOR]

As one of the important components of Earth observation technology, land use and land cover (LULC) image classification plays an essential role. It uses remote sensing techniques to classify specific categories of ground cover as a means of analyzing and understanding the natural attributes of the Earth's surface and the state of land use. It provides important information for applications in environmental protection, urban planning, and land resource management. However, remote sensing images are usually high-dimensional data and have limited available labeled samples, so performing the LULC classification task faces great challenges. In recent years, due to the emergence of deep learning technology, remote sensing data processing methods based on deep learning have achieved remarkable results, bringing new possibilities for the research and development of LULC classification. In this paper, we present a systematic review of deep-learning-based LULC classification, mainly covering the following five aspects: (1) introduction of the main components of five typical deep learning networks, how they work, and their unique benefits; (2) summary of two baseline datasets for LULC classification (pixel-level, patch-level) and performance metrics for evaluating different models (OA, AA, F1, and MIOU); (3) review of deep learning strategies in LULC classification studies, including convolutional neural networks (CNNs), autoencoders (AEs), generative adversarial networks (GANs), and recurrent neural networks (RNNs); (4) challenges faced by LULC classification and processing schemes under limited training samples; (5) outlooks on the future development of deep-learning-based LULC classification. [ABSTRACT FROM AUTHOR]

Deep learning has been successfully utilized in many applications, but it is vulnerable to adversarial samples. To address this vulnerability, a generative adversarial network (GAN) has been used to train a robust classifier. This paper presents a novel GAN model and its implementation to defend against L∞ and L2 constraint gradient-based adversarial attacks. The proposed model is inspired by some of the related work, but it includes multiple new designs such as a dual generator architecture, four new generator input formulations, and two unique implementations with L∞ and L2 norm constraint vector outputs. The new formulations and parameter settings of GAN are proposed and evaluated to address the limitations of adversarial training and defensive GAN training strategies, such as gradient masking and training complexity. Furthermore, the training epoch parameter has been evaluated to determine its effect on the overall training results. The experimental results indicate that the optimal formulation of GAN adversarial training must utilize more gradient information from the target classifier. The results also demonstrate that GANs can overcome gradient masking and produce effective perturbation to augment the data. The model can defend PGD L2 128/255 norm perturbation with over 60% accuracy and PGD L∞ 8/255 norm perturbation with around 45% accuracy. The results have also revealed that robustness can be transferred between the constraints of the proposed model. In addition, a robustness–accuracy tradeoff was discovered, along with overfitting and the generalization capabilities of the generator and classifier. These limitations and ideas for future work will be discussed. [ABSTRACT FROM AUTHOR]

For centuries, libraries worldwide have preserved ancient manuscripts due to their immense historical and cultural value. However, over time, both natural and human-made factors have led to the degradation of many ancient Arabic manuscripts, causing the loss of significant information, such as authorship, titles, or subjects, rendering them as unknown manuscripts. Although catalog cards attached to these manuscripts might contain some of the missing details, these cards have degraded significantly in quality over the decades within libraries. This paper presents a framework for identifying these unknown ancient Arabic manuscripts by processing the catalog cards associated with them. Given the challenges posed by the degradation of these cards, simple optical character recognition (OCR) is often insufficient. The proposed framework uses deep learning architecture to identify unknown manuscripts within a collection of ancient Arabic documents. This involves locating, extracting, and classifying the text from these catalog cards, along with implementing processes for region-of-interest identification, rotation correction, feature extraction, and classification. The results demonstrate the effectiveness of the proposed method, achieving an accuracy rate of 92.5%, compared to 83.5% with classical image classification and 81.5% with OCR alone. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) sensor often produces a shadow in pairs with the target due to its slant-viewing imaging. As a result, shadows in SAR images can provide critical discriminative features for classifiers, such as target contours and relative positions. However, shadows possess unique properties that differ from targets, such as low intensity and sensitivity to depression angles, making it challenging to extract depth features from shadows directly using convolutional neural networks (CNN). In this paper, we propose a new SAR image-classification framework to utilize target and shadow information comprehensively. First, we design a SAR image segmentation method to extract target regions and shadow masks. Second, based on SAR projection geometry, we propose a data-augmentation method to compensate for the geometric distortion of shadows due to differences in depression angles. Finally, we introduce a feature-enhancement module (FEM) based on depthwise separable convolution (DSC) and convolutional block attention module (CBAM), enabling deep networks to fuse target and shadow features adaptively. The experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset show that when only using target and shadow information, the published deep-learning models can still achieve state-of-the-art performance after embedding the FEM. [ABSTRACT FROM AUTHOR]

A global health emergency resulted from the COVID-19 epidemic. Image recognition techniques are a useful tool for limiting the spread of the pandemic; indeed, the World Health Organization (WHO) recommends the use of face masks in public places as a form of protection against contagion. Hence, innovative systems and algorithms were deployed to rapidly screen a large number of people with faces covered by masks. In this article, we analyze the current state of research and future directions in algorithms and systems for masked-face recognition. First, the paper discusses the importance and applications of facial and face mask recognition, introducing the main approaches. Afterward, we review the recent facial recognition frameworks and systems based on Convolution Neural Networks, deep learning, machine learning, and MobilNet techniques. In detail, we analyze and critically discuss recent scientific works and systems which employ machine learning (ML) and deep learning tools for promptly recognizing masked faces. Also, Internet of Things (IoT)-based sensors, implementing ML and DL algorithms, were described to keep track of the number of persons donning face masks and notify the proper authorities. Afterward, the main challenges and open issues that should be solved in future studies and systems are discussed. Finally, comparative analysis and discussion are reported, providing useful insights for outlining the next generation of face recognition systems. [ABSTRACT FROM AUTHOR]

Low-light images are a common phenomenon when taking photos in low-light environments with inappropriate camera equipment, leading to shortcomings such as low contrast, color distortion, uneven brightness, and high loss of detail. These shortcomings are not only subjectively annoying but also affect the performance of many computer vision systems. Enhanced low-light images can be better applied to image recognition, object detection and image segmentation. This paper proposes a novel RetinexDIP method to enhance images. Noise is considered as a factor in image decomposition using deep learning generative strategies. The involvement of noise makes the image more real, weakens the coupling relationship between the three components, avoids overfitting, and improves generalization. Extensive experiments demonstrate that our method outperforms existing methods qualitatively and quantitatively. [ABSTRACT FROM AUTHOR]

This paper considers the task of appearance-based localization: visual place recognition from omnidirectional images obtained from catadioptric cameras. The focus is on designing an efficient neural network architecture that accurately and reliably recognizes indoor scenes on distorted images from a catadioptric camera, even in self-similar environments with few discernible features. As the target application is the global localization of a low-cost service mobile robot, the proposed solutions are optimized toward being small-footprint models that provide real-time inference on edge devices, such as Nvidia Jetson. We compare several design choices for the neural network-based architecture of the localization system and then demonstrate that the best results are achieved with embeddings (global descriptors) yielded by exploiting transfer learning and fine tuning on a limited number of catadioptric images. We test our solutions on two small-scale datasets collected using different catadioptric cameras in the same office building. Next, we compare the performance of our system to state-of-the-art visual place recognition systems on the publicly available COLD Freiburg and Saarbrücken datasets that contain images collected under different lighting conditions. Our system compares favourably to the competitors both in terms of the accuracy of place recognition and the inference time, providing a cost- and energy-efficient means of appearance-based localization for an indoor service robot. [ABSTRACT FROM AUTHOR]

Recently, attention has been paid to the convolutional neural network (CNN) based synthetic aperture radar (SAR) target recognition method. Because of its advantages of automatic feature extraction and the preservation of translation invariance, the recognition accuracies are stronger than traditional methods. However, similar to other deep learning models, CNN is a "black-box" model, whose working process is vague. It is difficult to locate the decision reasons. Because of this, we focus on the process analysis of a pre-trained CNN model. The role of the processing to feature extraction and final recognition decision is discussed. The discussed components of CNN models are convolution, activation function, and full connection. Here, the convolution processing can be deemed as image filtering. The activation function provides a nonlinear element of processing. Moreover, the fully connected layers can also further extract features. In the experiment, four classical CNN models, i.e., AlexNet, VGG16, GoogLeNet, and ResNet-50, are trained by public MSTAR data, which can realize ten-category SAR target recognition. These pre-trained CNN models are processing objects of the proposed process analysis method. After the analysis, the content of the SAR image target features concerned by these pre-trained CNN models is further clarified. In summary, we provide a paradigm to process the analysis of pre-trained CNN models used for SAR target recognition in this paper. To some degree, the adaptability of these models to SAR images is verified. [ABSTRACT FROM AUTHOR]

While deep learning algorithms have advanced to a great extent, they are all designed for frame-based imagers that capture images at a high frame rate, which leads to a high storage requirement, heavy computations, and very high power consumption. Unlike frame-based imagers, event-based imagers output asynchronous pixel events without the need for global exposure time, therefore lowering both power consumption and latency. In this paper, we propose an innovative image recognition technique that operates on image events rather than frame-based data, paving the way for a new paradigm of recognizing objects prior to image acquisition. To the best of our knowledge, this is the first time such a concept is introduced featuring not only extreme early image recognition but also reduced computational overhead, storage requirement, and power consumption. Our collected event-based dataset using CeleX imager and five public event-based datasets are used to prove this concept, and the testing metrics reflect how early the neural network (NN) detects an image before the full-frame image is captured. It is demonstrated that, on average for all the datasets, the proposed technique recognizes an image 38.7 ms before the first perfect event and 603.4 ms before the last event is received, which is a reduction of 34% and 69% of the time needed, respectively. Further, less processing is required as the image is recognized 9460 events earlier, which is 37% less than waiting for the first perfectly recognized image. An enhanced NN method is also introduced to reduce this time. [ABSTRACT FROM AUTHOR]

Gaze estimation is an established research problem in computer vision. It has various applications in real life, from human–computer interactions to health care and virtual reality, making it more viable for the research community. Due to the significant success of deep learning techniques in other computer vision tasks—for example, image classification, object detection, object segmentation, and object tracking—deep learning-based gaze estimation has also received more attention in recent years. This paper uses a convolutional neural network (CNN) for person-specific gaze estimation. The person-specific gaze estimation utilizes a single model trained for one individual user, contrary to the commonly-used generalized models trained on multiple people's data. We utilized only low-quality images directly collected from a standard desktop webcam, so our method can be applied to any computer system equipped with such a camera without additional hardware requirements. First, we used the web camera to collect a dataset of face and eye images. Then, we tested different combinations of CNN parameters, including the learning and dropout rates. Our findings show that building a person-specific eye-tracking model produces better results with a selection of good hyperparameters when compared to universal models that are trained on multiple users' data. In particular, we achieved the best results for the left eye with 38.20 MAE (Mean Absolute Error) in pixels, the right eye with 36.01 MAE, both eyes combined with 51.18 MAE, and the whole face with 30.09 MAE, which is equivalent to approximately 1.45 degrees for the left eye, 1.37 degrees for the right eye, 1.98 degrees for both eyes combined, and 1.14 degrees for full-face images. [ABSTRACT FROM AUTHOR]

Hand gesture recognition from images is a critical task with various real-world applications, particularly in the field of human–robot interaction. Industrial environments, where non-verbal communication is preferred, are significant areas of application for gesture recognition. However, these environments are often unstructured and noisy, with complex and dynamic backgrounds, making accurate hand segmentation a challenging task. Currently, most solutions employ heavy preprocessing to segment the hand, followed by the application of deep learning models to classify the gestures. To address this challenge and develop a more robust and generalizable classification model, we propose a new form of domain adaptation using multi-loss training and contrastive learning. Our approach is particularly relevant in industrial collaborative scenarios, where hand segmentation is difficult and context-dependent. In this paper, we present an innovative solution that further challenges the existing approach by testing the model on an entirely unrelated dataset with different users. We use a dataset for training and validation and demonstrate that contrastive learning techniques in simultaneous multi-loss functions provide superior performance in hand gesture recognition compared to conventional approaches in similar conditions. [ABSTRACT FROM AUTHOR]

Due to the shortage of defect samples and the high cost of labelling during the process of hot-rolled strip production in the metallurgical industry, it is difficult to obtain a large quantity of defect data with diversity, which seriously affects the identification accuracy of different types of defects on the steel surface. To address the problem of insufficient defect sample data in the task of strip steel defect identification and classification, this paper proposes the Strip Steel Surface Defect-ConSinGAN (SDE-ConSinGAN) model for strip steel defect identification which is based on a single-image model trained by the generative adversarial network (GAN) and which builds a framework of image-feature cutting and splicing. The model aims to reduce training time by dynamically adjusting the number of iterations for different training stages. The detailed defect features of training samples are highlighted by introducing a new size-adjustment function and increasing the channel attention mechanism. In addition, real image features will be cut and synthesized to obtain new images with multiple defect features for training. The emergence of new images is able to richen generated samples. Eventually, the generated simulated samples can be directly used in deep-learning-based automatic classification of surface defects in cold-rolled thin strips. The experimental results show that, when SDE-ConSinGAN is used to enrich the image dataset, the generated defect images have higher quality and more diversity than the current methods do. [ABSTRACT FROM AUTHOR]

In recent years, different deep learning frameworks were introduced for hyperspectral image (HSI) classification. However, the proposed network models have a higher model complexity, and do not provide high classification accuracy if few-shot learning is used. This paper presents an HSI classification method that combines random patches network (RPNet) and recursive filtering (RF) to obtain informative deep features. The proposed method first convolves image bands with random patches to extract multi-level deep RPNet features. Thereafter, the RPNet feature set is subjected to dimension reduction through principal component analysis (PCA), and the extracted components are filtered using the RF procedure. Finally, the HSI spectral features and the obtained RPNet–RF features are combined to classify the HSI using a support vector machine (SVM) classifier. In order to test the performance of the proposed RPNet–RF method, some experiments were performed on three widely known datasets using a few training samples for each class, and classification results were compared with those obtained by other advanced HSI classification methods adopted for small training samples. The comparison showed that the RPNet–RF classification is characterized by higher values of such evaluation metrics as overall accuracy and Kappa coefficient. [ABSTRACT FROM AUTHOR]

This paper explored techniques for diagnosing breast cancer using deep learning based medical image recognition. X-ray (Mammography) images, ultrasound images, and histopathology images are used to improve the accuracy of the process by diagnosing breast cancer classification and by inferring their affected location. For this goal, the image recognition application strategies for the maximal diagnosis accuracy in each medical image data are investigated in terms of various image classification (VGGNet19, ResNet50, DenseNet121, EfficietNet v2), image segmentation (UNet, ResUNet++, DeepLab v3), and related loss functions (binary cross entropy, dice Loss, Tversky loss), and data augmentation. As a result of evaluations through the presented methods, when using filter-based data augmentation, ResNet50 showed the best performance in image classification, and UNet showed the best performance in both X-ray image and ultrasound image as image segmentation. When applying the proposed image recognition strategies for the maximal diagnosis accuracy in each medical image data, the accuracy can be improved by 33.3% in image segmentation in X-ray images, 29.9% in image segmentation in ultrasound images, and 22.8% in image classification in histopathology images. [ABSTRACT FROM AUTHOR]

In this study, we used image recognition technology to explore different ways to improve the safety of construction workers. Three object recognition scenarios were designed for safety at a construction site, and a corresponding object recognition model was developed for each scenario. The first object recognition model checks whether there are construction workers at the site. The second object recognition model assesses the risk of falling (falling off a structure or falling down) when working at an elevated position. The third object recognition model determines whether the workers are appropriately wearing safety helmets and vests. These three models were newly created using the image data collected from the construction sites and synthetic image data collected from the virtual environment based on transfer learning. In particular, we verified an artificial intelligence model based on a virtual environment in this study. Thus, simulating and performing tests on worker falls and fall injuries, which are difficult to re-enact by humans, are efficient algorithm verification methods. The verification and synthesis data acquisition method based on a virtual environment is one of the main contributions of this study. This paper describes the overall application development approach, including the structure and method used to collect the construction site image data, structure of the training image dataset, image dataset augmentation method, and the artificial intelligence backbone model applied for transfer learning. [ABSTRACT FROM AUTHOR]

Over recent years, with the advances in image recognition technology for deep learning, researchers have devoted continued efforts toward importing anomaly detection technology into the production line of automatic optical detection. Although unsupervised learning helps overcome the high costs associated with labeling, the accuracy of anomaly detection still needs to be improved. Accordingly, this paper proposes a novel deep learning model for anomaly detection to overcome this bottleneck. Leveraging a powerful pre-trained feature extractor and the skip connection, the proposed method achieves better feature extraction and image reconstructing capabilities. Results reveal that the areas under the curve (AUC) for the proposed method are higher than those of previous anomaly detection models for 16 out of 17 categories. This indicates that the proposed method can realize the most appropriate adjustments to the needs of production lines in order to maximize economic benefits. [ABSTRACT FROM AUTHOR]

Recently, laryngeal cancer cases have increased drastically across the globe. Accurate treatment for laryngeal cancer is intricate, especially in the later stages. This type of cancer is an intricate malignancy inside the head and neck area of patients. In recent years, diverse diagnosis approaches and tools have been developed by researchers for helping clinical experts to identify laryngeal cancer effectively. However, these existing tools and approaches have diverse issues related to performance constraints such as lower accuracy in the identification of laryngeal cancer in the initial stage, more computational complexity, and large time consumption in patient screening. In this paper, the authors present a novel and enhanced deep-learning-based Mask R-CNN model for the identification of laryngeal cancer and its related symptoms by utilizing diverse image datasets and CT images in real time. Furthermore, our suggested model is capable of capturing and detecting minor malignancies of the larynx portion in a significant and faster manner in the real-time screening of patients, and it saves time for the clinicians, allowing for more patient screening every day. The outcome of the suggested model is enhanced and pragmatic and obtained an accuracy of 98.99%, precision of 98.99%, F1 score of 97.99%, and recall of 96.79% on the ImageNet dataset. Several studies have been performed in recent years on laryngeal cancer detection by using diverse approaches from researchers. For the future, there are vigorous opportunities for further research to investigate new approaches for laryngeal cancer detection by utilizing diverse and large dataset images. [ABSTRACT FROM AUTHOR]

Thermal cameras, as opposed to RBG cameras, work effectively in extremely low illumination situations and can record data outside of the human visual spectrum. For surveillance and security applications, thermal images have several benefits. However, due to the little visual information in thermal images and intrinsic similarity of facial heat maps, completing face identification tasks in the thermal realm is particularly difficult. It can be difficult to attempt identification across modalities, such as when trying to identify a face in thermal images using the ground truth database for the matching visible light domain or vice versa. We proposed a method for detecting objects and actions on thermal human face images, based on the classification of five different features (hat, glasses, rotation, normal, and hat with glasses) in this paper. This model is presented in five steps. To improve the results of feature extraction during the pre-processing step, initially, we resize the images and then convert them to grayscale level using a median filter. In addition, features are extracted from pre-processed images using principle component analysis (PCA). Furthermore, the horse herd optimization algorithm (HOA) is employed for feature selection. Then, to detect the human face in thermal images, the LeNet-5 method is used. It is utilized to detect objects and actions in face areas. Finally, we classify the objects and actions on faces using the ANU-Net approach with the Monarch butterfly optimization (MBO) algorithm to achieve higher classification accuracy. According to experiments using the Terravic Facial Infrared Database, the proposed method outperforms "state-of-the-art" methods for face recognition in thermal images. Additionally, the results for several facial recognition tasks demonstrate good precision. [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]

In this paper, I propose a bird eye view image detection method for parking areas and collision risk areas at the same time in parking situations. Deep learning algorithms using area detection and semantic segmentation were used. The main architecture of the method described in this paper is based on a harmonic densely connected network and a cross-stage partial network. The dataset used for training was calibrated to four 190° wide-angle cameras to generate around view monitor (AVM) images based on the Chungbuk National University parking lot, and an experiment was performed based on this dataset. In the experimental results, the available parking area was visualized by detecting the parking line, parking area, and available driving area in the AVM images. Furthermore, the undetected area in the semantic segmentation as a collision risk area was visualized in order to obtain the results. According to the proposed attention CSPHarDNet model, the experimental results were 81.89% mIoU and 18.36 FPS in a NVIDIA Xavier environment. The results of this experiment demonstrated that algorithms can be used in real time in a parking situation and have better performance results compared to the conventional HarDNet. [ABSTRACT FROM AUTHOR]

Recently, the machine-vision-based blades surface damage detection technique has received great attention for its low cost, easy operation, and lack of a need for prior knowledge. The rapid progress of deep learning has contributed to the promotion of this technology with automatic feature extraction, a broader scope of application, and stronger expansibility. An image recognition method of wind turbine blade defects using attention-based MobileNetv1-YOLOv4 and transfer learning is proposed in this paper. The backbone convolution neural network of YOLOv4 is replaced by the lightweight MobileNetv1 for feature extraction to reduce complexity and computation. Attention-based feature refinement with three distinctive modules, SENet, ECANet, and CBAM, is introduced to realize adaptive feature optimization. To solve the problem of slow network convergence and low detection accuracy caused by insufficient data, a two-stage transfer learning approach is introduced to fine-tune the pre-trained network. Comparative experiments verify the efficacy of the proposed model, with higher detection accuracy but a significantly faster response speed and less computational complexity, compared with other state-of-the-art networks by using images of the wind turbine blades taken by an unmanned aerial vehicle (UAV). A sensitivity study is also conducted to present the effects of different training dataset sizes on the model performance. [ABSTRACT FROM AUTHOR]

Breast cancer is the second most common cancer worldwide, primarily affecting women, while histopathological image analysis is one of the possibile methods used to determine tumor malignancy. Regarding image analysis, the application of deep learning has become increasingly prevalent in recent years. However, a significant issue is the unbalanced nature of available datasets, with some classes having more images than others, which may impact the performance of the models due to poorer generalizability. A possible strategy to avoid this problem is downsampling the class with the most images to create a balanced dataset. Nevertheless, this approach is not recommended for small datasets as it can lead to poor model performance. Instead, techniques such as data augmentation are traditionally used to address this issue. These techniques apply simple transformations such as translation or rotation to the images to increase variability in the dataset. Another possibility is using generative adversarial networks (GANs), which can generate images from a relatively small training set. This work aims to enhance model performance in classifying histopathological images by applying data augmentation using GANs instead of traditional techniques. [ABSTRACT FROM AUTHOR]

Cloudy conditions at a local scale pose a significant challenge for forecasting renewable energy generation through photovoltaic panels. Consequently, having real-time knowledge of sky conditions becomes highly valuable. This information could inform decision-making processes in system operations, such as determining whether conditions are favorable for activating a standalone system requiring a minimum level of radiation or whether sky conditions might lead to higher energy consumption than generation during adverse cloudy conditions. This research leveraged convolutional neural networks (CNNs) and transfer learning (TL) classification techniques, testing various architectures from the EfficientNet family and two ResNet models for classifying sky images. Cross-validation methods were applied across different experiments, where the most favorable outcome was achieved with the EfficientNetV2-B1 and EfficientNetV2-B2 models boasting a mean Accuracy of 98.09%. This study underscores the efficacy of the architectures employed for sky image classification, while also highlighting the models yielding the best results. [ABSTRACT FROM AUTHOR]

Most logit-based knowledge distillation methods transfer soft labels from the teacher model to the student model via Kullback–Leibler divergence based on softmax, an exponential normalization function. However, this exponential nature of softmax tends to prioritize the largest class (target class) while neglecting smaller ones (non-target classes), leading to an oversight of the non-target classes's significance. To address this issue, we propose Non-Target-Class-Enhanced Knowledge Distillation (NTCE-KD) to amplify the role of non-target classes both in terms of magnitude and diversity. Specifically, we present a magnitude-enhanced Kullback–Leibler (MKL) divergence multi-shrinking the target class to enhance the impact of non-target classes in terms of magnitude. Additionally, to enrich the diversity of non-target classes, we introduce a diversity-based data augmentation strategy (DDA), further enhancing overall performance. Extensive experimental results on the CIFAR-100 and ImageNet-1k datasets demonstrate that non-target classes are of great significance and that our method achieves state-of-the-art performance across a wide range of teacher–student pairs. [ABSTRACT FROM AUTHOR]

Malware classification is a crucial step in defending against potential malware attacks. Despite the significance of a robust malware classifier, existing approaches reveal notable limitations in achieving high performance in malware classification. This study focuses on image-based malware detection, where malware binaries are transformed into visual representations to leverage image classification techniques. We propose a two-branch deep network designed to capture salient features from these malware images. The proposed network integrates faster asymmetric spatial attention to refine the extracted features of its backbone. Additionally, it incorporates an auxiliary feature branch to learn missing information about malware images. The feasibility of the proposed method has been thoroughly examined and compared with state-of-the-art deep learning-based classification methods. The experimental results demonstrate that the proposed method can surpass its counterparts across various evaluation metrics. [ABSTRACT FROM AUTHOR]

Inspections of concrete bridges across the United States represent a significant commitment of resources, given their biannual mandate for many structures. With a notable number of aging bridges, there is an imperative need to enhance the efficiency of these inspections. This study harnessed the power of computer vision to streamline the inspection process. Our experiment examined the efficacy of a state-of-the-art Visual Transformer (ViT) model combined with distinct image enhancement detector algorithms. We benchmarked against a deep learning Convolutional Neural Network (CNN) model. These models were applied to over 20,000 high-quality images from the Concrete Images for Classification dataset. Traditional crack detection methods often fall short due to their heavy reliance on time and resources. This research pioneers bridge inspection by integrating ViT with diverse image enhancement detectors, significantly improving concrete crack detection accuracy. Notably, a custom-built CNN achieves over 99% accuracy with substantially lower training time than ViT, making it an efficient solution for enhancing safety and resource conservation in infrastructure management. These advancements enhance safety by enabling reliable detection and timely maintenance, but they also align with Industry 4.0 objectives, automating manual inspections, reducing costs, and advancing technological integration in public infrastructure management. [ABSTRACT FROM AUTHOR]

Fine-grained representation is fundamental to species classification based on deep learning, and in this context, cross-modal contrastive learning is an effective method. The diversity of species coupled with the inherent contextual ambiguity of natural language poses a primary challenge in the cross-modal representation alignment of conservation area image data. Integrating cross-modal retrieval tasks with generation tasks contributes to cross-modal representation alignment based on contextual understanding. However, during the contrastive learning process, apart from learning the differences in the data itself, a pair of encoders inevitably learns the differences caused by encoder fluctuations. The latter leads to convergence shortcuts, resulting in poor representation quality and an inaccurate reflection of the similarity relationships between samples in the original dataset within the shared space of features. To achieve fine-grained cross-modal representation alignment, we first propose a residual attention network to enhance consistency during momentum updates in cross-modal encoders. Building upon this, we propose momentum encoding from a multi-task perspective as a bridge for cross-modal information, effectively improving cross-modal mutual information, representation quality, and optimizing the distribution of feature points within the cross-modal shared semantic space. By acquiring momentum encoding queues for cross-modal semantic understanding through multi-tasking, we align ambiguous natural language representations around the invariant image features of factual information, alleviating contextual ambiguity and enhancing model robustness. Experimental validation shows that our proposed multi-task perspective of cross-modal momentum encoders outperforms similar models on standardized image classification tasks and image–text cross-modal retrieval tasks on public datasets by up to 8% on the leaderboard, demonstrating the effectiveness of the proposed method. Qualitative experiments on our self-built conservation area image–text paired dataset show that our proposed method accurately performs cross-modal retrieval and generation tasks among 8142 species, proving its effectiveness on fine-grained cross-modal image–text conservation area image datasets. [ABSTRACT FROM AUTHOR]

The calculation of land surface temperatures (LSTs) via low-altitude thermal infrared remote (TIR) sensing images at a block scale is gaining attention. However, the accurate calculation of LSTs requires a precise determination of the range of various underlying surfaces in the TIR images, and existing approaches face challenges in effectively segmenting the underlying surfaces in the TIR images. To address this challenge, this study proposes a deep learning (DL) methodology to complete the instance segmentation and quantification of underlying surfaces through the low-altitude TIR image dataset. Mask region-based convolutional neural networks were utilized for pixel-level classification and segmentation with an image dataset of 1350 annotated TIR images of an urban rail transit hub with a complex distribution of underlying surfaces. Subsequently, the hyper-parameters and architecture were optimized for the precise classification of the underlying surfaces. The algorithms were validated using 150 new TIR images, and four evaluation indictors demonstrated that the optimized algorithm outperformed the other algorithms. High-quality segmented masks of the underlying surfaces were generated, and the area of each instance was obtained by counting the true-positive pixels with values of 1. This research promotes the accurate calculation of LSTs based on the low-altitude TIR sensing images. [ABSTRACT FROM AUTHOR]

Face recognition has been well studied under visible light and infrared (IR) in both intra-spectral and cross-spectral cases. However, how to fuse different light bands for face recognition, i.e., hyperspectral face recognition, is still an open research problem, which has the advantages of richer information retention and all-weather functionality over single-band face recognition. Thus, in this research, we revisit the hyperspectral recognition problem and provide a deep learning-based approach. A new fusion model (named HyperFace) is proposed to address this problem. The proposed model features a pre-fusion scheme, a Siamese encoder with bi-scope residual dense learning, a feedback-style decoder, and a recognition-oriented composite loss function. Experiments demonstrate that our method yields a much higher recognition rate than face recognition using only visible light or IR data. Moreover, our fusion model is shown to be superior to other general-purpose image fusion methods that are either traditional or deep learning-based, including state-of-the-art methods, in terms of both image quality and recognition performance. [ABSTRACT FROM AUTHOR]

Activity recognition based on inertial sensors is an essential task in mobile and ubiquitous computing. To date, the best performing approaches in this task are based on deep learning models. Although the performance of the approaches has been increasingly improving, a number of issues still remain. Specifically, in this paper we focus on the issue of the dependence of today's state-of-the-art approaches to complex ad hoc deep learning convolutional neural networks (CNNs), recurrent neural networks (RNNs), or a combination of both, which require specialized knowledge and considerable effort for their construction and optimal tuning. To address this issue, in this paper we propose an approach that automatically transforms the inertial sensors time-series data into images that represent in pixel form patterns found over time, allowing even a simple CNN to outperform complex ad hoc deep learning models that combine RNNs and CNNs for activity recognition. We conducted an extensive evaluation considering seven benchmark datasets that are among the most relevant in activity recognition. Our results demonstrate that our approach is able to outperform the state of the art in all cases, based on image representations that are generated through a process that is easy to implement, modify, and extend further, without the need of developing complex deep learning models. [ABSTRACT FROM AUTHOR]

Currently, the target detection based on convolutional neural network plays an important role in image recognition, speech recognition and other fields. However, the current network model features a complex structure, a huge number of parameters and resources. These conditions make it difficult to apply in embedded devices with limited computational capabilities and extreme sensitivity to power consumption. In this regard, the application scenarios of deep learning are limited. This paper proposes a real-time detection scheme for cook assistant overalls based on the Hi3559A embedded processor. With YOLOv3 as the benchmark network, this scheme fully mobilizes the hardware acceleration resources through the network model optimization and the parallel processing technology of the processor, and improves the network reasoning speed, so that the embedded device can complete the task of real-time detection on the local device. The experimental results show that through the purposeful cropping, segmentation and in-depth optimization of the neural network according to the specific processor, the neural network can recognize the image accurately. In an application environment where the power consumption is only 5.5 W, the recognition speed of the neural network on the embedded end is increased to about 28 frames (the design requirement was to achieve a recognition speed of 25 frames or more), so that the optimized network can be effectively applied in the back kitchen overalls identification scene. [ABSTRACT FROM AUTHOR]

Coronavirus disease 2019 (COVID-19), originating in China, has rapidly spread worldwide. Physicians must examine infected patients and make timely decisions to isolate them. However, completing these processes is difficult due to limited time and availability of expert radiologists, as well as limitations of the reverse-transcription polymerase chain reaction (RT-PCR) method. Deep learning, a sophisticated machine learning technique, leverages radiological imaging modalities for disease diagnosis and image classification tasks. Previous research on COVID-19 classification has encountered several limitations, including binary classification methods, single-feature modalities, small public datasets, and reliance on CT diagnostic processes. Additionally, studies have often utilized a flat structure, disregarding the hierarchical structure of pneumonia classification. This study aims to overcome these limitations by identifying pneumonia caused by COVID-19, distinguishing it from other types of pneumonia and healthy lungs using chest X-ray (CXR) images and related tabular medical data, and demonstrate the value of incorporating tabular medical data in achieving more accurate diagnoses. Resnet-based and VGG-based pre-trained convolutional neural network (CNN) models were employed to extract features, which were then combined using early fusion for the classification of eight distinct classes. We leveraged the hierarchal structure of pneumonia classification within our approach to achieve improved classification outcomes. Since an imbalanced dataset is common in this field, a variety of versions of generative adversarial networks (GANs) were used to generate synthetic data. The proposed approach tested in our private datasets of 4523 patients achieved a macro-avg F1-score of 95.9% and an F1-score of 87.5% for COVID-19 identification using a Resnet-based structure. In conclusion, in this study, we were able to create an accurate deep learning multi-modal to diagnose COVID-19 and differentiate it from other kinds of pneumonia and normal lungs, which will enhance the radiological diagnostic process. [ABSTRACT FROM AUTHOR]

ECG classification or heartbeat classification is an extremely valuable tool in cardiology. Deep learning-based techniques for the analysis of ECG signals assist human experts in the timely diagnosis of cardiac diseases and help save precious lives. This research aims at digitizing a dataset of images of ECG records into time series signals and then applying deep learning (DL) techniques on the digitized dataset. State-of-the-art DL techniques are proposed for the classification of the ECG signals into different cardiac classes. Multiple DL models, including a convolutional neural network (CNN), a long short-term memory (LSTM) network, and a self-supervised learning (SSL)-based model using autoencoders are explored and compared in this study. The models are trained on the dataset generated from ECG plots of patients from various healthcare institutes in Pakistan. First, the ECG images are digitized, segmenting the lead II heartbeats, and then the digitized signals are passed to the proposed deep learning models for classification. Among the different DL models used in this study, the proposed CNN model achieves the highest accuracy of ∼92%. The proposed model is highly accurate and provides fast inference for real-time and direct monitoring of ECG signals that are captured from the electrodes (sensors) placed on different parts of the body. Using the digitized form of ECG signals instead of images for the classification of cardiac arrhythmia allows cardiologists to utilize DL models directly on ECG signals from an ECG machine for the real-time and accurate monitoring of ECGs. [ABSTRACT FROM AUTHOR]

Three-dimensional (3D) range-gated imaging can obtain high spatial resolution intensity images as well as pixel-wise depth information. Several algorithms have been developed to recover depth from gated images such as the range-intensity correlation algorithm and deep-learning-based algorithm. The traditional range-intensity correlation algorithm requires specific range-intensity profiles, which are hard to generate, while the existing deep-learning-based algorithm requires large number of real-scene training data. In this work, we propose a method of range-intensity-profile-guided gated light ranging and imaging to recover depth from gated images based on a convolutional neural network. In this method, the range-intensity profile (RIP) of a given gated light ranging and imaging system is obtained to generate synthetic training data from Grand Theft Auto V for our range-intensity ratio and semantic network (RIRS-net). The RIRS-net is mainly trained on synthetic data and fine-tuned with RIP data. The network learns both semantic depth cues and range-intensity depth cues in the synthetic data, and learns accurate range-intensity depth cues in the RIP data. In the evaluation experiments on both a real-scene and synthetic test dataset, our method shows a better result compared to other algorithms. [ABSTRACT FROM AUTHOR]

In this work, an exhaustive analysis of the partial discharges that originate in the bubbles present in dielectric mineral oils is carried out. To achieve this, a low-cost, high-resolution CMOS image sensor is used. Partial discharge measurements using that image sensor are validated by a standard electrical detection system that uses a discharge capacitor. In order to accurately identify the images corresponding to partial discharges, a convolutional neural network is trained using a large set of images captured by the image sensor. An image classification model is also developed using deep learning with a convolutional network based on a TensorFlow and Keras model. The classification results of the experiments show that the accuracy achieved by our model is around 95% on the validation set and 82% on the test set. As a result of this work, a non-destructive diagnosis method has been developed that is based on the use of an image sensor and the design of a convolutional neural network. This approach allows us to obtain information about the state of mineral oils before breakdown occurs, providing a valuable tool for the evaluation and maintenance of these dielectric oils. [ABSTRACT FROM AUTHOR]

Remote sensing image classification (RSIC) is designed to assign specific semantic labels to aerial images, which is significant and fundamental in many applications. In recent years, substantial work has been conducted on RSIC with the help of deep learning models. Even though these models have greatly enhanced the performance of RSIC, the issues of diversity in the same class and similarity between different classes in remote sensing images remain huge challenges for RSIC. To solve these problems, a duplex-hierarchy representation learning (DHRL) method is proposed. The proposed DHRL method aims to explore duplex-hierarchy spaces, including a common space and a label space, to learn discriminative representations for RSIC. The proposed DHRL method consists of three main steps: First, paired images are fed to a pretrained ResNet network for extracting the corresponding features. Second, the extracted features are further explored and mapped into a common space for reducing the intra-class scatter and enlarging the inter-class separation. Third, the obtained representations are used to predict the categories of the input images, and the discrimination loss in the label space is minimized to further promote the learning of discriminative representations. Meanwhile, a confusion score is computed and added to the classification loss for guiding the discriminative representation learning via backpropagation. The comprehensive experimental results show that the proposed method is superior to the existing state-of-the-art methods on two challenging remote sensing image scene datasets, demonstrating that the proposed method is significantly effective. [ABSTRACT FROM AUTHOR]

The number of wheat ears is an essential indicator for wheat production and yield estimation, but accurately obtaining wheat ears requires expensive manual cost and labor time. Meanwhile, the characteristics of wheat ears provide less information, and the color is consistent with the background, which can be challenging to obtain the number of wheat ears required. In this paper, the performance of Faster regions with convolutional neural networks (Faster R-CNN) and RetinaNet to predict the number of wheat ears for wheat at different growth stages under different conditions is investigated. The results show that using the Global WHEAT dataset for recognition, the RetinaNet method, and the Faster R-CNN method achieve an average accuracy of 0.82 and 0.72, with the RetinaNet method obtaining the highest recognition accuracy. Secondly, using the collected image data for recognition, the R2 of RetinaNet and Faster R-CNN after transfer learning is 0.9722 and 0.8702, respectively, indicating that the recognition accuracy of the RetinaNet method is higher on different data sets. We also tested wheat ears at both the filling and maturity stages; our proposed method has proven to be very robust (the R2 is above 90). This study provides technical support and a reference for automatic wheat ear recognition and yield estimation. [ABSTRACT FROM AUTHOR]

Berry production is increasing worldwide each year; however, high production leads to labor shortages and an increase in wasted fruit during harvest seasons. This problem opened new research opportunities in computer vision as one main challenge to address is the uncontrolled light conditions in greenhouses and open fields. The high light variations between zones can lead to underexposure of the regions of interest, making it difficult to classify between vegetation, ripe, and unripe blackberries due to their black color. Therefore, the aim of this work is to automate the process of classifying the ripeness stages of blackberries in normal and low-light conditions by exploring the use of image fusion methods to improve the quality of the input image before the inference process. The proposed algorithm adds information from three sources: visible, an improved version of the visible, and a sensor that captures images in the near-infrared spectra, obtaining a mean F1 score of 0.909 ± 0.074 and 0.962 ± 0.028 in underexposed images, without and with model fine-tuning, respectively, which in some cases is an increase of up to 12% in the classification rates. Furthermore, the analysis of the fusion metrics showed that the method could be used in outdoor images to enhance their quality; the weighted fusion helps to improve only underexposed vegetation, improving the contrast of objects in the image without significant changes in saturation and colorfulness. [ABSTRACT FROM AUTHOR]

Coronavirus 2019 (COVID-19) has posed a serious threat to the lives and health of the majority of people worldwide. Since the early days of the outbreak, South Korea's government and citizens have made persistent efforts to provide effective prevention against further spread of the disease. In particular, the participation of individual citizens in complying with the necessary code of conduct to prevent spread of the infection, through measures such as social distancing and mask wearing, is as instrumental as the geographical tracking of the trajectory of the infected. In this paper, we propose an activity recognition method based on a wristband equipped with an IR array and inertial measurement unit (IMU) to detect individual compliance with codes of personal hygiene management, such as mask wearing, which are recommended to prevent the spread of infectious diseases. The results of activity recognition were comparatively analyzed by applying conventional machine learning algorithms and convolutional neural networks (CNNs) to the IMU time series and IR array thermal images collected from 25 subjects. When CNN and 24 × 32 thermal images were used, 97.8% accuracy was achieved (best performance), and when 6 × 8 low-resolution thermal images were used, similar performance with 97.1% accuracy was obtained. In the case of using IMU, the performance of activity recognition was lower than that obtained with the IR array, but an accuracy of 93% was achieved even in the case of applying machine learning algorithms, indicating that it is more suitable for wearable devices with low computational capability. [ABSTRACT FROM AUTHOR]

The lack of labeled training samples restricts the improvement of Hyperspectral Remote Sensing Image (HRSI) classification accuracy based on deep learning methods. In order to improve the HRSI classification accuracy when there are few training samples, a Lightweight 3D Dense Autoencoder Network (L3DDAN) is proposed. Structurally, the L3DDAN is designed as a stacked autoencoder which consists of an encoder and a decoder. The encoder is a hybrid combination of 3D convolutional operations and 3D dense block for extracting deep features from raw data. The decoder composed of 3D deconvolution operations is designed to reconstruct data. The L3DDAN is trained by unsupervised learning without labeled samples and supervised learning with a small number of labeled samples, successively. The network composed of the fine-tuned encoder and trained classifier is used for classification tasks. The extensive comparative experiments on three benchmark HRSI datasets demonstrate that the proposed framework with fewer trainable parameters can maintain superior performance to the other eight state-of-the-art algorithms when there are only a few training samples. The proposed L3DDAN can be applied to HRSI classification tasks, such as vegetation classification. Future work mainly focuses on training time reduction and applications on more real-world datasets. [ABSTRACT FROM AUTHOR]

Sooty mold is a common disease found in citrus plants and is characterized by black fungi growth on fruits, leaves, and branches. This mold reduces the plant's ability to carry out photosynthesis. In small leaves, it is very difficult to detect sooty mold at the early stages. Deep learning-based image recognition techniques have the potential to identify and diagnose pest damage and diseases such as sooty mold. Recent studies used advanced and expensive hyperspectral or multispectral cameras attached to UAVs to examine the canopy of the plants and mid-range cameras to capture close-up infected leaf images. To bridge the gap on capturing canopy level images using affordable camera sensors, this study used a low-cost home surveillance camera to monitor and detect sooty mold infection on citrus canopy combined with deep learning algorithms. To overcome the challenges posed by varying light conditions, the main reason for using specialized cameras, images were collected at night, utilizing the camera's built-in night vision feature. A total of 4200 sliced night-captured images were used for training, 200 for validation, and 100 for testing, employed on the YOLOv5m, YOLOv7, and CenterNet models for comparison. The results showed that YOLOv7 was the most accurate in detecting sooty molds at night, with 74.4% mAP compared to YOLOv5m (72%) and CenterNet (70.3%). The models were also tested using preprocessed (unsliced) night images and day-captured sliced and unsliced images. The testing on preprocessed (unsliced) night images demonstrated the same trend as the training results, with YOLOv7 performing best compared to YOLOv5m and CenterNet. In contrast, testing on the day-captured images had underwhelming outcomes for both sliced and unsliced images. In general, YOLOv7 performed best in detecting sooty mold infections at night on citrus canopy and showed promising potential in real-time orchard disease monitoring and detection. Moreover, this study demonstrated that utilizing a cost-effective surveillance camera and deep learning algorithms can accurately detect sooty molds at night, enabling growers to effectively monitor and identify occurrences of the disease at the canopy level. [ABSTRACT FROM AUTHOR]

Regarding the interpretable techniques in the field of image recognition, Grad-CAM is widely used for feature localization in images to reflect the logical decision-making information behind the neural network due to its high applicability. However, extensive experimentation on a customized dataset revealed that the deep convolutional neural network (CNN) model based on Gradient-weighted Class Activation Mapping (Grad-CAM) technology cannot effectively resist the interference of large-scale noise. In this article, an optimization of the deep CNN model was proposed by incorporating the Dropkey and Dropout (as a comparison) algorithm. Compared with Grad-CAM, the improved Grad-CAM based on Dropkey applies an attention mechanism to the feature map before calculating the gradient, which can introduce randomness and eliminate some areas by applying a mask to the attention score. Experimental results show that the optimized Grad-CAM deep CNN model based on the Dropkey algorithm can effectively resist large-scale noise interference and achieve accurate localization of image features. For instance, under the interference of a noise variance of 0.6, the Dropkey-enhanced ResNet50 model achieves a confidence level of 0.878 in predicting results, while the other two models exhibit confidence levels of 0.766 and 0.481, respectively. Moreover, it exhibits excellent performance in visualizing tasks related to image features such as distortion, low contrast, and small object characteristics. Furthermore, it has promising prospects in practical computer vision applications. For instance, in the field of autonomous driving, it can assist in verifying whether deep learning models accurately understand and process crucial objects, road signs, pedestrians, or other elements in the environment. [ABSTRACT FROM AUTHOR]