Search

Smart farming plays a revolutionary paradigm shift in the realm of agriculture to enhance farming practices across the globe. Traditional smart farming methods face various challenges such as inefficiency, resource wastage, weather dependency, and environmental impact. To overcome these challenges, this paper proposed a novel method named Densely Connected Convolutional based Golden search (DCC-GS) with initial Search strategy for smart farming in agriculture. In this study, a densely connected convolutional network is utilized to capture intricate patterns in the data which is significant for smart farming. In this work, the Golden search optimization algorithm is utilized for hyperparameters tuning and enhancing the effectiveness of the DCC-GS method. Also, an initial search strategy is implemented to decrease the computational burden of the DCC-GS method. The DCC-GS method is validated on the Crop Recommendation dataset. The performance evaluation of the DCC-GS method involves analyzing its effectiveness based on key measures such as cost, specificity, recall, F1-score, precision, and accuracy and comparing its performance against existing methods to assess the DCC-GS method's effectiveness. The DCC-GS method achieved accuracy, precision, recall, F1-score, and specificity of 98.63%, 97.95%, 97.83%, 97.86%, and 97.82% respectively. The experimental evaluation illustrates the effectiveness of the DCC-GS method for smart farming in agriculture. Moreover, Recognizing the importance of agriculture in global food security, many farmers are motivated to adopt smart farming practices to contribute to stable and sufficient food supplies. [ABSTRACT FROM AUTHOR]

Image classification, the primary domain where deep neural networks significantly contribute to image analysis, requires a substantial amount of computer memory to train. This is particularly true in the fully connected layer, which accounts for 90% of the total memory. Moreover, the flattening operation could potentially result in the loss of the multi-linear structure of the image data. The tensor regression network, however, minimally impacts the performance of the neural network while achieving a high compression rate. This effectively mitigates the issue of large memory occupation in the neural network model. The DenseNet model, in particular, can alleviate the vanishing-gradient problem and strengthen feature propagation and outperform other existing networks. This article proposes a novel tensor network model that embeds the tensor regression layer into the DenseNet model. The framework of this tensor DenseNet model has been established, and its estimation procedure is developed. Tensor network model is applied to the classification of the following datasets: Fruits 360, 100 Sports Image, ASL Alphabet, and Mini-ImageNet. The experimental results indicate that the combination of the DenseNet model with the tensor regression layer not only conserves a significant amount of memory but also maintains a high accuracy of classification, compared with existing tensor network models. [ABSTRACT FROM AUTHOR]

Vascular Dementia is a severe disease that results from dead nerve cells’ accumulation in blood vessels. This affects the blood flow and impairs memory and decision-making abilities. Machine learning and deep learning have been used in detecting this disease. Nevertheless, their accuracy has been inconsistent, explaining why their utilization in diagnosing patients has led to poor performance. We developed several transfer learning architectures that improve classification accuracy and diagnosis performance in assessing vascular dementia. The process first entails the preprocessing of the dataset where a random selection ensures data representation is balanced. We used a dataset containing resting-state fMRI scans to split training, testing, and validation into 80%, 10%, and 10%. We employ different Convolutional Neural Network architectures such as VGG16, VGG19, DenseNet121, and InceptionResNetV2 to enhance classification. To, enhance these, we incorporated Rectified Linear Unit and leaky activation functions for the training phase to counteract problems associated with vanishing gradient common in deep learning tasks. Our methodology ensures effective information flow throughout different layers, which is essential for a divergent information hierarchy in medical information. As such, the specifics show that our approach achieved 84.67% in accuracy in the multi-classification, which is better than the current state-of-the-art research in the same field. Therefore, the result shows that transfer learning-based approaches are suitable when combined with strategic pre-processing and activation functions in improving the diagnosis of vascular dementia using MRI images.

The stratigraphic correlation of well logs plays an essential role in characterizing subsurface reservoirs. However, it suffers from a small amount of training data and expensive computing time. In this work, we propose the Attention Based Dense Network (ASDNet) for the stratigraphic correlation of well logs. To implement the suggested model, we first employ the attention mechanism to the input well logs, which can effectively generate the weighted well logs to serve for further feature extraction. Subsequently, the DenseNet is utilized to achieve good feature reuse and avoid gradient vanishing. After model training, we employ the ASDNet to the testing data set and evaluate its performance based on the well log data set from Northwest China. Finally, the numerical results demonstrate that the suggested ASDNet provides higher prediction accuracy for automated stratigraphic correlation of well logs than state-of-the-art contrastive UNet and SegNet.

In the audio recognition field, environmental sound classification (ESC) is more arduous when compared with music information recognition (MIR) and automatic speech recognition (ASR) featuring structured sounds. To make full use of both channel and spatial characteristics of Log-Mel spectrogram extracted from the environmental sound, a spatial-channel attention mechanism built on densely connected convolutional network (DenseNet) is proposed. First, a DenseNet is proposed to obtain the feature extraction of Log-Mel spectrogram. A spatial-channel attention mechanism that can excavate both the channel and spatial correlation is utilized to enable the proposed network to focus more on prominent features. Next, to address the over-fitting caused by data scarcity, a mix-up data augmentation method is applied to Log-Mel spectrogram, ensuring the diversity of data. Last, the effectiveness of the proposed method is verified on two public datasets (ESC-50 and ESC-10). Results show the proposed spatial-channel attention mechanism allows the network to increase the recognition accuracy of environmental sounds to 79. 3% (ESC-50) and 94. 3% (ESC-10) respectively. [ABSTRACT FROM AUTHOR]

We present a pocket-size densely connected convolutional network (DenseNet) directed to classification of size-normalized colour images according to varieties of grapes captured in those images. We compare the DenseNet with three established small-size networks in terms of performance, inference time and model size. We propose a data augmentation that we use in training the networks. We train and evaluate the networks on in-field images. The trained networks distinguish between seven grapevine varieties and background, where four and three varieties, respectively, are of red and green grapes. Compared to the established networks, the DenseNet is characterized by near state-of-the-art performance, short inference time and minimal model size. All these aspects qualify the network for real-time, mobile and edge computing applications. The DenseNet opens possibilities for constructing affordable selective harvesters in accordance with agriculture 4.0. [ABSTRACT FROM AUTHOR]

The fault diagnosis methods of rolling bearing based on model and signal processing and analysis have the problems of modeling difficulty and signal characteristic extraction difficulty. The rolling bearing fault diagnosis method based on shallow machine learning has limited capability to learn the characteristics of complex data. The convolutional neural networks are often used in rolling bearing fault diagnosis methods based on deep learning. But with the deepening of the network, gradient dispersion or disappearance will occur. And directly converting the rolling bearing vibration signal into one-dimensional or two-dimensional images as network input will not preserve the time correlation between the signals, resulting in the loss of signal information. To solve these problems, a fault diagnosis method for rolling bearing based on Markov transition field(MTF) and densely connected convolutional networks(DenseNet) is proposed. The vibration signal of the rolling bearing is coded by MTF to generate a two-dimensional image. The time sequence information and the state transition information of the signal are preserved. The two-dimensional image is taken as the input of DenseNet, and the fault characteristics of the rolling bearing vibration signal are extracted through DenseNet. The method enhances the propagation of characteristic information, makes full use of characteristic information, and then realizes fault classification and identification. The data on the Case Western Reserve University bearing dataset is used for the test. The results show that the method can effectively identify the fault types of rolling bearings, and the accuracy of fault diagnosis is 99.5%. In order to further verify the fault diagnosis capability and superiority of this method when the motor load changes, four kinds of network input images, namely, gray-scale image, envelope spectrum image, cepstrum image and MTF generation image, are selected for comparative experiments with the method of combining three networks, namely, Inception, ResNet and DenseNet. The results show that the fault diagnosis accuracy of different methods is higher when the motor load is unchanged than when the motor load is changed. The fault diagnosis accuracy of MTF+DenseNet method is higher than that of other methods. The proposed method still has a high fault diagnosis accuracy when the motor load changes, with an average value of 94.53% and good generalization performance.

With the grid-connected application of renewable energy sources such as wind and photovoltaic power, the nonlinearity and fluctuation of load data makes load forecasting more difficult than ever before. In order to extract the implicit relationship between multiple features and power load to construct a long-term sequence dependency, this paper proposes a short-term load forecasting based on improved temporal convolutional network (TCN) and densely connected convolutional network (DenseNet). Firstly, multiple features are reconstructed by using a fixed-length sliding window, and then the high-dimensional features reflecting the complex and non-stationary characteristics of power load are extracted by the DenseNet to construct a feature matrix. Secondly, we innovatively improve the TCN and introduce a parallel pooling into the traditional TCN to mine the features of time sequences. Finally, the self-attention mechanism (SAM) is used to further enhance the weight of key features to eliminate the influences of interference signals. Experiments were performed on Southern China and ISO-NE (New England) public datasets to verify the effectiveness and generalization of the proposed model. Compared with the traditional TCN, the mean average percentage error (MAPE) of the improved TCN on the two datasets decreases by 23.38% and 8.14%, respectively. Furthermore, when compared to the TCN-SAM hybrid model, the MAPE of the proposed model is significantly reduced by 42.41% and 26.89%, respectively.

The surface defect detection (SDD) problem is one of the crucial techniques during production process, so it has become a key research area to control the quality of industrial products, which has been increasingly of greater interest to the researchers especially with the rapid development of artificial neural networks technology in recent years. Therefore, this paper proposes a novel deep convolutional neural network algorithm aiming at SDD. Firstly, a dense cross-stage partial Darknet backbone network is designed for feature extraction by optimizing cross-stage partial Darknet through the idea of dense connections, which can, not only enhance feature reuse but also greatly alleviate the overfitting issue. Secondly, a new cross-stage hierarchy module is presented combining the cross-stage feature fusion strategy and depthwise separable convolution technique for each node of the path aggregated feature pyramid network (PAN). Finally, an efficient channel attention (ECA) mechanism is introduced in PAN to construct a novel ECA PAN. The experimental results on three surface defect datasets show that the mean average precision of this network is 2.63, 5.48, and 1.16% which is higher than that of the baseline network, respectively. The proposed network outperforms not only the classical models but state-of-the-art models, which indicates the proposed algorithm can achieve higher accuracy and speed with fewer calculation parameters. And what is more, the proposed algorithm also has outstanding generalization ability. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Mine Automation is the property of Industry & Mine Automation Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The model-based and signal-based rolling bearing fault diagnosis methods have problems such as difficult modeling and cumbersome signal analysis. The data-driven rolling bearing fault diagnosis methods mostly use convolutional neural networks, but as the number of network layers increases during network training, gradient disappearance occurs. Moreover, taking the vibration signal of the rolling bearing directly as the network input will cause incomplete feature extraction. In order to solve the above problems, a rolling bearing fault diagnosis method based on Gramian angular field(GAF) and densely connected convolutional network(DenseNet) is proposed. The one-dimensional time series of rolling bearing vibration signals are converted into two-dimensional images by GAF, which preserves the correlation information between the time series data. The two-dimensional images are used as the input of the densely connected convolutional network, and the feature extraction of the two-dimensional images is carried out by the DenseNet, which improves the feature information utilization and realizes the fault classification. Experiments are carried out by using the data from the Case Western Reserve University bearing dataset. The results show that the method can identify rolling bearing fault types effectively with a fault diagnosis accuracy rate of 99.75%. In order to further prove the superiority of this method, the fault diagnosis methods of gray-scale image+DenseNet, GAF+residual network(ResNet), gray-scale image + ResNet are selected for comparison. The results show that the GAF+DenseNet method has the highest accuracy rate, and the gray-scale image+ResNet method has the lowest accuracy rate. Compared with the gray-scale image, the GAF converted two-dimensional image retains the relevant information between the original time series data. Compared with ResNet, DenseNet is able to extract the fault features more adequately due to denser connection method.

Deep neural networks (DNNs) have recently been significantly applied to automatic arrhythmia classification. However, their classification accuracy still has room for improvement.The aim of this study is to address the existing limitations in current models by developing a more effective approach for automatic arrhythmia classification. The specific objectives include enhancing the receptive field sizes to capture more detailed information across various temporal scales, and incorporating inter-channel correlations to improve the feature extraction process.This study proposes a pyramidal dense connectivity layer and bidirectional long short-term memory network (PDC-BiLSTM) to effectively extract waveform features across various temporal scales, which can capture the intricate details and the broader global information in the signals through a wide range of sensory fields. The efficient channel attention (ECA) is additionally introduced to dynamically allocate weights to each feature channel, assisting the model inefficiently prioritizing essential characteristics during the training process.The experimental results on the MIT-BIH arrhythmia database showed that the overall classification accuracy of the proposed method under the intra-patient paradigm reached 99.82%, and the positive predictive value, sensitivity and F1 Score were 99.64%, 97.61% and 98.60% respectively; under the inter-patient paradigm, the overall accuracy was 96.30%.Compared with the latest research results in this field, the proposed model is also better than the existing models in terms of accuracy, which has the potential value of being applied to devices that assist in diagnosing cardiovascular diseases. [ABSTRACT FROM AUTHOR]

Developing a colorimetry-based artificial scent screening system (i.e., an olfactory visual sensing system) with high sensitivity and accurate pattern recognition for detecting fruit ripeness remains challenging. In this work, we construct a flexible dye/CelluMOFs-based sensor array with improved sensitivity for on-site detection of characteristic gases of fruits and integrate a densely connected convolutional network (DenseNet) into the sensor array, enabling it to recognize unique scent fingerprints and categorize the ripeness of fruits. In the system, CelluMOFs are synthesized through in situ growth of γ-cyclodextrin metal-organic frameworks (γ-CD-MOFs) on flexible fiber filter paper to fabricate a uniform, flexible and porous dye/CelluMOFs sensitive membrane. Compared to the pristine filter paper, the CelluMOFs exhibit increased porosity with a 62 times higher specific surface area and a 3-fold increase in dye loading capacity after 12 h of adsorption. The prepared dye/CelluMOFs sensing film shows outstanding mechanical and detection stability with negligible deviation after 100 cycles of rubbing. The colorimetric visualization arrays with multiple colorimetric dye/CelluMOFs chips, enable the sensitive recognition and detection of nine kinds of characteristic fruit odors and achieve a high response at 8-1500 ppm of trans -2-hexenal, showcasing remarkably low gas detection thresholds. On the basis of the ppm-level limit of detection with high sensitivity, the fabricated colorimetric sensor arrays are typically used for in situ assessment of fruit ripeness by integrating DenseNet. This approach achieves a satisfactory classification accuracy of 99.09% on the validation set, enabling high-precision prediction of fruit ripeness levels.

Purpose: It plays a significant role to accurately and automatically segment lesions from ultrasound (US) images in clinical application. Nevertheless, it is extremely challenging because distinct components of heterogeneous lesions are similar to background in US images. In our study, a transfer learning-based method is developed for full-automatic joint segmentation of nodular lesions. Methods: Transfer learning is a widely used method to build high performing computer vision models. Our transfer learning model is a novel type of densely connected convolutional network (SDenseNet). Specifically, we pre-train SDenseNet based on ImageNet dataset. Then our SDenseNet is designed as a multi-channel model (denoted Mul-DenseNet) for automatically jointly segmenting lesions. As comparison, our SDenseNet using different transfer learning is applied to segmenting nodules, respectively. In our study, we find that more datasets for pre-training and multiple pre-training do not always work in segmentation of nodules, and the performance of transfer learning depends on a judicious choice of dataset and characteristics of targets. Results: Experimental results illustrate a significant performance of the Mul-DenseNet compared to that of other methods in the study. Specially, for thyroid nodule segmentation, overlap metric (OM), dice ratio (DR), true-positive rate (TPR), false-positive rate (FPR) and modified Hausdorff distance (MHD) are 0.9257 ± 0.0027 , 0.9596 ± 0.0009 , 0.9869 ± 0.0008 , 0.0183 ± 0.0066 and 0.3897 ± 0.3488 mm, respectively; for breast nodule segmentation, OM, DR, TPR, FPR and MHD are 0.8912 ± 0.0072 , 0.9513 ± 0.0038 , 0.9835 ± 0.0015 , 0.2381 ± 0.1301 and 0.2017 ± 0.0302 mm, respectively. Conclusions: The experimental results illustrate our transfer learning models are very effective in segmentation of lesions, which also demonstrate that it is potential of our proposed Mul-DenseNet model in clinical applications. This model can reduce heavy workload of the physicians so that it can avoid misdiagnosis cases due to excessive fatigue. Moreover, it is easy and reproducible to detect lesions without medical expertise. [ABSTRACT FROM AUTHOR]

Deep residual network (ResNet) had shown remarkable performance in image recognition tasks, due to its shorter connections between layers close to the input and those close to the output. And densely connected convolutional network (DenseNet) had further improved recognition performance by dense feature reuses. To improve the performance of residual units in ResNet and feature reuses in DenseNet, we propose a simple and effective convolutional network architecture, named double reuses based residual network (DRRNet). DRRNet improves the residual unit of ResNet, where the feature reuse connections are added to combine all feature maps from the convolutional layers to produce the residual, and uses a residual reuse path outside units to reuse all residuals as the final feature maps for classification. Residual learning used in DRRNet can alleviate the vanishing-gradient problem. The double reuses including inner-unit feature reuses and outer-unit residual reuses effectively decrease computational cost as compared with dense connections in DenseNet, and further strengthen the forward feature propagation. DRRNet is evaluated on three object recognition benchmark datasets and an object detection dataset. In comparison with the state-of-the-art, DRRNet achieves a good balance between classification accuracy and computational cost, and optimal detection performance. [Display omitted] • Our network improves the residual units, inside which the features are reused. • Our network reuses all residuals in the residual reuse path outside the units. • The output of residual reuse path is used as the final feature maps. • Our network strengthens the forward feature propagation by using double reuses. • Our network gets a good balance between classification accuracy and calculation cost. [ABSTRACT FROM AUTHOR]

The fusion of infrared and visible images can utilize the indication characteristics and the textural details of source images to realize the all-weather detection. The deep learning (DL) based fusion solutions can reduce the computational cost and complexity compared with traditional methods since there is no need to design complex feature extraction methods and fusion rules. There are no standard reference images and the publicly available infrared and visible image pairs are scarce. Most supervised DL-based solutions have to take pre-training on other labeled large datasets which may not behave well when testing. The few unsupervised fusion methods can hardly obtain ideal images with good visual impression. In this paper, an infrared and visible image fusion method based on unsupervised convolutional neural network is proposed. When designing the network structure, densely connected convolutional network (DenseNet) is used as the sub-network for feature extraction and reconstruction to ensure that more information of source images can be retained in the fusion images. As to loss function, the perceptual loss is creatively introduced and combined with the structure similarity loss to constrain the updating of weight parameters during the back propagation. The perceptual loss designed helps to improve the visual information fidelity (VIF) of the fusion image effectively. Experimental results show that this method can obtain fusion images with prominent targets and obvious details. Compared with other 7 traditional and deep learning methods, the fusion results of this method are better on objective evaluation and visual observation when taken together.

Firearm violence is one of the leading causes of death in many countries around the world, including Thailand. This work proposes a fast and accurate automated method to classify firearm brands from bullet markings. Specifically, a panoramic image of a bullet collected from a crime scene was captured using a developed mobile phone application and custom-built portable hardware. The top three state-of-the-art CNNs pretrained on ImageNet-DenseNet121, ResNet50, and Xception-were further trained on the same training set, which was composed of 718 bullets collected from eight different firearm brands-Beretta, Browning, CZ, Glock, Norinco, Ruger, Sig Sauer, and Smith & Wesson-using a five-fold cross validation technique. DenseNet121 provided the highest AUC of 0.99 for CZ classification (the most common registered firearm brand in Thailand) and the highest average AUC for the eight firearm brands (0.9780 ± 0.0130 SD), which was significantly higher than those of ResNet50 and Xception. In addition, there were no interaction effects between the CNN model and firearm brand on AUC. DenseNet121, which had the highest AUC, was evaluated on the test set (72 bullets), and the results showed that the Beretta and CZ classifications had the lowest accuracy (91.18%), followed by the Browning and Norinco classifications (96.88%), whereas the Glock, Ruger, Sig Sauer, and Smith & Wesson classifications had the highest accuracy (98.41%). These results suggest that the developed mobile phone application based on a deep learning algorithm and the custom-built portable hardware have promising potential for use at crime scenes to classify firearms from bullet markings. By narrowing down the list of suspects, this convenient approach can potentially accelerate bullet identification processes for many forensic science examiners.

Copyright of Journal of Mine Automation is the property of Industry & Mine Automation Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Remote Sensing is the property of Editorial Office of Journal of Remote Sensing & Science Publishing Co. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Breast mass is one of the most distinctive signs for the diagnosis of breast cancer, and the accurate segmentation of masses is critical for improving the accuracy of breast cancer detection and reducing the mortality rate. It is time-consuming for a physician to review the film. Besides, traditional medical segmentation techniques often require prior knowledge or manual extraction of features, which often lead to a subjective diagnosis. Therefore, developing an automatic image segmentation method is important for clinical application. In this paper, a fully automatic method based on deep learning for breast mass segmentation is proposed, which combines densely connected U-Net with attention gates (AGs). It contains an encoder and a decoder. The encoder is a densely connected convolutional network and the decoder is the decoder of U-Net integrated with AGs. The proposed method is tested on the public and authoritative database-Digital Database for Screening Mammography (DDSM) database. F1-score, mean intersection over union, sensitivity, specificity, and overall accuracy are used to evaluate the effectiveness of the proposed method. The experimental results show that dense U-Net integrated AGs achieve better segmentation results than U-Net, attention U-Net, DenseNet, and state-of-the-art methods.

With the rapid progress of urbanization, predicting citywide crowd flows has become increasingly significant in many fields, such as traffic management and public security. However, influenced by the complex spatiotemporal relations in raw data and other factors, such as events and weather, obtaining a precise prediction is challenging. Some previous works attempted to address this problem using various ways, such as autoregressive integrated moving average, vector auto-regression and some deep learning models. However, seldom can these methods comprehensively capture the spatiotemporal correlations. In this paper, we propose a novel spatio-temporal prediction model that is based on densely connected convolutional networks and attention long short-term memory (ST-DCCNAL), to simultaneously predict the inflow and outflow of the crowds in regions divided within a specific city. The ST-DCCNAL model consists of three parts: spatial part, external factors part and temporal part. In the spatial part, we employ densely connected convolutional networks to extract spatial characteristics at different levels. The external factors part utilizes a fully connected network to extract features from auxiliary information. In the last part, an attention-based long short-term memory module is leveraged to capture the temporal pattern. To demonstrate the practicality and effectiveness of the proposed model, we evaluate it using two separate real-world datasets of taxis in Beijing and bikes in New York. The experimental results confirm that the performance of our model is better than that of other baseline methods.

• 3D fatigue crack quantification based on a densely connect convolutional network-based regression model. • Fusion of linear and nonlinear features of ultrasonic guided wave for fatigue crack quantification. • Versatility in both early awareness and long-term assessment of 3D fatigue cracks. • Validation using experimental data. Linear and nonlinear ultrasonic guided wave (UGW)-based methods are available for fatigue crack detection and quantification. However, linear methods usually fail to detect undersized cracks, while nonlinear methods struggle to assess cracks once they have extended to a certain degree. The fusion of linear and nonlinear UGW features offers a new opportunity to enhance evaluation precision and effectiveness, yet it necessitates highly complex modeling. Motivated by this, a regression model is proposed based on the famed densely connected convolutional network, DenseNet, for three-dimensional (3D) fatigue crack quantification in both crack initiation and growth stages. Via the continuous wavelet transform (CWT), the spectra of UGW signals embracing both linear and nonlinear features of UGW are obtained. Subsequently, DenseNet is adopted to extract implicit features of spectra images. Finally, the last fully connected layer of DenseNet is modified as a regression layer to estimate the length and depth of 3D fatigue cracks. A dataset comprising 500 UGW signals is created for validating the proposed model. The results demonstrate that the model can characterize the length and depth of 3D cracks in both initiation and growth stages, establishing it as a promising proof-of-concept for the deep learning-based quantification of 3D fatigue cracks. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel densely connected convolutional module (DCCM)-based convolutional neural network for human pose estimation, which can achieve higher parameter efficiency compared to the state-of-the-art works. Although existing methods for human pose estimation have achieved considerable accuracy, the number of required model parameters and computation complexity are relatively high. To solve this problem, we propose to use a DCCM as the basic unit of the neural network. For each layer of DCCM, feature maps that all preceding layers produce are concatenated as its input, and its own output feature maps are delivered to each subsequent layer. The experimental results on the MPII human pose data set and LSP data set show that our method can get comparable performance, while it requires less parameters, which means higher parameter efficiency can be achieved. Furthermore, we explore that how different configurations of the proposed network structure can affect the accuracy of human pose estimation.

Recently, there has been a huge rise in malware growth, which creates a significant security threat to organizations and individuals. Despite the incessant efforts of cybersecurity research to defend against malware threats, malware developers discover new ways to evade these defense techniques. Traditional static and dynamic analysis methods are ineffective in identifying new malware and pose high overhead in terms of memory and time. Typical machine learning approaches that train a classifier based on handcrafted features are also not sufficiently potent against these evasive techniques and require more efforts due to feature-engineering. Recent malware detectors indicate performance degradation due to class imbalance in malware datasets. To resolve these challenges, this work adopts a visualization-based method, where malware binaries are depicted as two-dimensional images and classified by a deep learning model. We propose an efficient malware detection system based on deep learning. The system uses a reweighted class-balanced loss function in the final classification layer of the DenseNet model to achieve significant performance improvements in classifying malware by handling imbalanced data issues. Comprehensive experiments performed on four benchmark malware datasets show that the proposed approach can detect new malware samples with higher accuracy (98.23% for the Malimg dataset, 98.46% for the BIG 2015 dataset, 98.21% for the MaleVis dataset, and 89.48% for the unseen Malicia dataset) and reduced false-positive rates when compared with conventional malware mitigation techniques while maintaining low computational time. The proposed malware detection solution is also reliable and effective against obfuscation attacks.

Delineating lesion boundaries play a central role in diagnosing thyroid and breast cancers, making related therapy plans and evaluating therapeutic effects. However, it is often time-consuming and error-prone with limited reproducibility to manually annotate low-quality ultrasound (US) images, given high speckle noises, heterogeneous appearances, ambiguous boundaries etc., especially for nodular lesions with huge intra-class variance. It is hence appreciative but challenging for accurate lesion segmentations from US images in clinical practices. In this study, we propose a new densely connected convolutional network (called MDenseNet) architecture to automatically segment nodular lesions from 2D US images, which is first pre-trained over ImageNet database (called PMDenseNet) and then retrained upon the given US image datasets. Moreover, we also designed a deep MDenseNet with pre-training strategy (PDMDenseNet) for segmentation of thyroid and breast nodules by adding a dense block to increase the depth of our MDenseNet. Extensive experiments demonstrate that the proposed MDenseNet-based method can accurately extract multiple nodular lesions, with even complex shapes, from input thyroid and breast US images. Moreover, additional experiments show that the introduced MDenseNet-based method also outperforms three state-of-the-art convolutional neural networks in terms of accuracy and reproducibility. Meanwhile, promising results in nodular lesion segmentation from thyroid and breast US images illustrate its great potential in many other clinical segmentation tasks., Competing Interests: Declaration of competing interest No conflicts of interest., (Copyright © 2023 Elsevier Ltd. All rights reserved.)

Convolutional neural network (CNN) architectures have shown excellent image classification performance on large‐scale visual recognition tasks. If a CNN architecture contains a shorter connection between layers close to the input and those close to the output, the training can be deeper, more accurate and efficient. In this Lette, the authors propose a densely connected CNN architecture for facial expression recognition (FER‐Net), which connects the output of each convolution layer to the inputs of the next convolution layers in the architecture. Experiments conducted on a publicly available dataset show that FER‐Net produces state‐of‐the‐art results in facial expression recognition. [ABSTRACT FROM AUTHOR]

• Deep learning methods detect cells forming human iPS cell in early reprogramming stage. • CNN classifies a brightfield microscopy image as probability images of cell classes with specific features. • U-net segments cells on a microscopy image and classifies to count the cells by its probability images for the early detection. • RNN by training the probability images predicts morphology of future human iPS cell colonies. • Test for 150 sets of time-lapse images from reprogramming CD34+ cells achieved high detection accuracy. Human induced pluripotent stem cells (hiPSCs) represent an ideal source for patient specific cell-based regenerative medicine; however, efficiency of hiPSC formation from reprogramming cells is low. We use several deep-learning results from time-lapse brightfield microscopy images during culture, to early detect the cells potentially reprogramming into hiPSCs and predict the colony morphology of these cells for improving efficiency of culturing a new hiPSC line. Sets of time-lapse bright-field images are taken to track reprogramming process of CD34+ cells biologically identified as just beginning reprogramming. Prior the experiment, 9 classes of templates with distinct cell features clipped from microscopy images at various reprogramming stages are used to train a CNN model. The CNN is then used to classify a microscopy image as probability images of these classes. Probability images of some class are used to train a densely connected convolutional network for extracting regions of this class on a microscopy image. A U-net is trained to segment cells on the time-lapse images in early reprogramming stage during culture. The segmented cells are classified by the extracted regions to count various types of cells appearing in the early reprogramming stage for predicting the identified cells potentially forming hiPSCs. The probability images of hiPSC classes are also used to train a spatiotemporal RNN for predicting the future hiPSC colony morphology of the potential cells. Experimental results show the prediction (before 7 days after of beginning of the reprogramming) achieved 0.8 accuracy, and 66% of the identified cells under different culture conditions, predicted as forming, finally formed hiPSCs. The predicted hiPSC images and extracted colonies on the images show the prediction for future 1.5 days achieved high accuracy of hiPSC colony areas and image similarity. Our study proposes a method using several deep learning models to efficiently select the reprogramming cells possibly forming hiPSCs and to predict the shapes of growing hiPSC colonies. The proposed method is expected to improve the efficiency when establishing a new hiPSC line culture. [ABSTRACT FROM AUTHOR]

Recently, there has been a huge rise in malware growth, which creates a significant security threat to organizations and individuals. Despite the incessant efforts of cybersecurity research to defend against malware threats, malware developers discover new ways to evade these defense techniques. Traditional static and dynamic analysis methods are ineffective in identifying new malware and pose high overhead in terms of memory and time. Typical machine learning approaches that train a classifier based on handcrafted features are also not sufficiently potent against these evasive techniques and require more efforts due to feature-engineering. Recent malware detectors indicate performance degradation due to class imbalance in malware datasets. To resolve these challenges, this work adopts a visualization-based method, where malware binaries are depicted as two-dimensional images and classified by a deep learning model. We propose an efficient malware detection system based on deep learning. The system uses a reweighted class-balanced loss function in the final classification layer of the DenseNet model to achieve significant performance improvements in classifying malware by handling imbalanced data issues. Comprehensive experiments performed on four benchmark malware datasets show that the proposed approach can detect new malware samples with higher accuracy (98.23% for the Malimg dataset, 98.46% for the BIG 2015 dataset, 98.21% for the MaleVis dataset, and 89.48% for the unseen Malicia dataset) and reduced false-positive rates when compared with conventional malware mitigation techniques while maintaining low computational time. The proposed malware detection solution is also reliable and effective against obfuscation attacks. [ABSTRACT FROM AUTHOR]

• We converted EEG time series into graphs and then used densely connected convolutional network to identify epilepsy. • We implemented signal segmentation and voting strategy to solve the classification of long-term EEG signals. • Our method achieves 100 % classification accuracy in all six commonly used epilepsy EEG classification experiments. The classification of epileptiform electroencephalogram (EEG) signals has been treated as an important but challenging issue for realizing epileptic seizure detection. In this work, combing gray recurrence plot (GRP) and densely connected convolutional network (DenseNet), we developed a novel classification system named GRP-DNet to identify seizures and epilepsy from single-channel, long-term EEG signals. The proposed GRP-DNet classification system includes three main modules: 1) input module takes an input long-term EEG signal and divides it into multiple short segments using a fixed-size non-overlapping sliding window (FNSW); 2) conversion module transforms short segments into GRPs and passes them to the DenseNet; 3) fusion and decision, the predicted label of each GRP is fused using a majority voting strategy to make the final decision. Six different classification experiments were designed based on a publicly available benchmark database to evaluate the effectiveness of our system. Experimental results showed that the proposed GRP-DNet achieved an excellent classification accuracy of 100 % in each classification experiment, Furthermore, GRP-DNet gave excellent computational efficiency, which indicates its tremendous potential for developing an EEG-based online epilepsy diagnosis system. Our GRP-DNet system was superior to the existing competitive classification systems using the same database. The GRP-DNet is a potentially powerful system for identifying and classifying EEG signals recorded from different brain states. [ABSTRACT FROM AUTHOR]

Objective: To train convolutional networks using multi-lead ECG data and classify new data accurately to provide reliable information for clinical diagnosis., Methods: The data were pre-processed with a bandpass filter, and signal framing was adopted to adjust the data of different lengths to the same size to facilitate network training and prediction. The dataset was expanded by increasing the sample size to improve the detection rate of abnormal samples. A depth-wise separable convolution structure was used for more specific feature extraction for different channels of twelve-lead ECG data. We trained the two classifiers for each label using the improved DenseNet to classify different labels., Results: The propose model showed an accuracy of 80.13% for distinguishing between normal and abnormal ECG with a sensitivity of 80.38%, a specificity of 79.91% and a F1 score of 79.35%., Conclusions: The model proposed herein can rapidly and effectively classify the ECG data. The running time of a single dataset on GPU is 33.59 ms, which allows real-time prediction to meet the clinical requirements.