Convolutional Neural Network Method Research Articles

Bridging Convolutional Neural Networks and Transformers for Efficient Crack Detection in Concrete Building Structures.

Detecting cracks in building structures is an essential practice that ensures safety, promotes longevity, and maintains the economic value of the built environment. In the past, machine learning (ML) and deep learning (DL) techniques have been used to enhance classification accuracy. However, the conventional CNN (convolutional neural network) methods incur high computational costs owing to their extensive number of trainable parameters and tend to extract only high-dimensional shallow features that may not comprehensively represent crack characteristics. We proposed a novel convolution and composite attention transformer network (CCTNet) model to address these issues. CCTNet enhances crack identification by processing more input pixels and combining convolution channel attention with window-based self-attention mechanisms. This dual approach aims to leverage the localized feature extraction capabilities of CNNs with the global contextual understanding afforded by self-attention mechanisms. Additionally, we applied an improved cross-attention module within CCTNet to increase the interaction and integration of features across adjacent windows. The performance of CCTNet on the Historical Building Crack2019, SDTNET2018, and proposed DS3 has a precision of 98.60%, 98.93%, and 99.33%, respectively. Furthermore, the training validation loss of the proposed model is close to zero. In addition, the AUC (area under the curve) is 0.99 and 0.98 for the Historical Building Crack2019 and SDTNET2018, respectively. CCTNet not only outperforms existing methodologies but also sets a new standard for the accurate, efficient, and reliable detection of cracks in building structures.

Identification of Potato Plant Pests Using the Convolutional Neural Network VGG16 Method

Pests are one of the main challenges in potato cultivation that can significantly reduce crop yields. Therefore, quick and accurate pest identification is crucial for effective pest control. This research aims to develop a pest identification system for potato plants using the Convolutional Neural Network (CNN) method with the VGG16 architecture. The dataset used consists of images of pests commonly found on potato plants. After the labeling process, these images were used to train the CNN VGG16 model. The research results show that the CNN VGG16 method can identify types of pests with an accuracy rate of 73%. The results serve as a reference to help farmers and agricultural practitioners detect the presence of pests earlier and take the necessary actions to reduce crop losses.

Comparative analysis of potato blight diseases BARI-72 and BARI-73 using a simplified convolutional neural network method

Comparative analysis of potato blight diseases BARI-72 and BARI-73 using a simplified convolutional neural network method

Pengembangan Aplikasi Mobile Android untuk Deteksi Otomatis Mata Katarak Menggunakan CNN dan Tensorflow

The development of an Android mobile application for automatic cataract detection using Convolutional Neural Network (CNN) and TensorFlow has been conducted. The aim of this research is to provide an easily accessible solution for the public to detect cataracts early, thereby reducing the negative impact of this eye condition. The CNN method is utilized to recognize cataract patterns through image data, with TensorFlow serving as the primary development platform. Preprocessing steps and data processing are implemented to enhance the detection accuracy and address variations in eye images. Evaluation indicates that the application is capable of detecting cataracts with satisfactory accuracy, making it a potential tool for cataract prevention and early management. In conclusion, this application enables rapid and efficient cataract detection, improving the accessibility of eye care and contributing to enhancing overall quality of life for communities by providing early intervention and treatment options.

Ground Coverage Classification in UAV Image Using a Convolutional Neural Network Feature Map

Background: To understand land transformation at the local level, there is a need to develop new strategies appropriate for land management policies and practices. In various geographical research, ground coverage plays an important role particularly in planning, physical geography explorations, environmental analysis, and sustainable planning. Objective: The research aimed to analyze land cover using vegetation density data collected through remote sensing. Specifically, the data assisted in land processing and land cover classification based on vegetation density. Methods: Before classification, image was preprocessed using Convolutional Neural Network (CNN) architecture's ResNet 50 and DenseNet 121 feature extraction methods. Furthermore, several algorithm were used, namely Decision Tree, Naí¯ve Bayes, K-Nearest Neighbor, Random Forest, Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost). Results: Classification comparison between methods showed that using CNN method obtained better results than machine learning. By using CNN architecture for feature extraction, SVM method, which adopted ResNet-50 for feature extraction, achieved an impressive accuracy of 85%. Similarly using SVM method with DenseNet121 feature extraction led to a performance of 81%. Conclusion: Based on results comparing CNN and machine learning, ResNet 50 architecture performed the best, achieving a result of 92%. Meanwhile, SVM performed better than other machine learning method, achieving an 84% accuracy rate with ResNet-50 feature extraction. XGBoost came next, with an 82% accuracy rate using the same ResNet-50 feature extraction. Finally, SVM and XGBoost produced the best results for feature extraction using DenseNet-121, with an accuracy rate of 81%. Keywords: Classification, CNN Architecture, Feature Extraction, Ground Coverage, Vegetation Density.

Researching the CNN Collaborative Inference Mechanism for Heterogeneous Edge Devices.

Convolutional Neural Networks (CNNs) have been widely applied in various edge computing devices based on intelligent sensors. However, due to the high computational demands of CNN tasks, the limited computing resources of edge intelligent terminal devices, and significant architectural differences among these devices, it is challenging for edge devices to independently execute inference tasks locally. Collaborative inference among edge terminal devices can effectively utilize idle computing and storage resources and optimize latency characteristics, thus significantly addressing the challenges posed by the computational intensity of CNNs. This paper targets efficient collaborative execution of CNN inference tasks among heterogeneous and resource-constrained edge terminal devices. We propose a pre-partitioning deployment method for CNNs based on critical operator layers, and optimize the system bottleneck latency during pipeline parallelism using data compression, queuing, and "micro-shifting" techniques. Experimental results demonstrate that our method achieves significant acceleration in CNN inference within heterogeneous environments, improving performance by 71.6% compared to existing popular frameworks.

E-Nose: Time-Frequency Attention Convolutional Neural Network for Gas Classification and Concentration Prediction.

In the electronic nose (E-nose) systems, gas type recognition and accurate concentration prediction are some of the most challenging issues. This study introduced an innovative pattern recognition method of time-frequency attention convolutional neural network (TFA-CNN). A time-frequency attention block was designed in the network, aiming to excavate and effectively integrate the temporal and frequency domain information in the E-nose signals to enhance the performance of gas classification and concentration prediction tasks. Additionally, a novel data augmentation strategy was developed, manipulating the feature channels and time dimensions to reduce the interference of sensor drift and redundant information, thereby enhancing the model's robustness and adaptability. Utilizing two types of metal-oxide-semiconductor gas sensors, this research conducted qualitative and quantitative analysis on five target gases. The evaluation results showed that the classification accuracy could reach 100%, and the coefficient of the determination (R2) score of the regression task was up to 0.99. The Pearson correlation coefficient (r) was 0.99, and the mean absolute error (MAE) was 1.54 ppm. The experimental test results were almost consistent with the system predictions, and the MAE was 1.39 ppm. This study provides a method of network learning that combines time-frequency domain information, exhibiting high performance in gas classification and concentration prediction within the E-nose system.

Progress in the application of artificial intelligence-assisted molecular modification of enzymes

Natural enzymes are often difficult to meet the needs of application and research in terms of activity, enantiomer selectivity or thermal stability. Therefore, it is an important task of enzyme engineering to explore efficient molecular modification technologies to improve the properties of such enzymes. The molecular modification technologies of enzymes mainly include rational design, directed evolution, and artificial intelligence-assisted design. Directed evolution and rational design are experiment-driven molecular modification approaches of enzymes and have been successfully applied to enzyme engineering. However, due to the huge space sizes of protein sequences and the lack of experimental data, the current modification methods still face major challenges. With the development of next-generation sequencing, high-throughput screening, protein databases, and artificial intelligence (AI), data-driven enzyme engineering is emerging as a promising solution to these challenges. The AI-assisted statistical learning method has been used to establish a model for predicting the sequence/structure-properties of enzymes in a data-driven manner. Excellent mutant enzymes can be selected according to the prediction results, which greatly improve the efficiency of molecular modification. Considering the application requirements of molecular modification of enzymes, this paper reviews the data acquisition methods and application examples of AI-assisted molecular modification of enzymes, with focuses on the convolutional neural network method for predicting protein thermostability, aiming to provide reference for researchers in this field.

Optimized convolution neural network-based model for detection and classification of pulmonary diseases

Pelican Optimization Algorithm-based Convolutional Neural Network (POA-CNN) method for the automated identification of pulmonary disorders such as COVID-19 and pneumonia is proposed in this research. The method makes use of several processing layers in order to comprehend the representation of stratified data. The three primary phases of the model are feature extraction via POA-based hyperparameter optimization, image classification, and image pre-processing. This approach improves existing systems' performance in detecting pulmonary diseases, highlighting the potential of deep learning in identifying and categorizing human diseases. The study uses a resizing, grayscale, and augmentation method to optimize an existing CNN model. A Convolutional Neural Network (CNN) is then applied to classify Pneumonia and Covid-19 cases. The proposed model achieves an accuracy rate of 97.28 and 97.00%, outperforming existing models. This technique is effective in detecting and classifying other pulmonary diseases, and can be used to automatically detect and classify these diseases. Higher accuracy findings show how successful the model is, making it a useful tool for pulmonary illness identification.

Acute leukemia prediction and classification using convolutional neural network and generative adversarial network

Acute Leukemia (AL) is a type of cancer that affects the blood cells and can be fatal if not detected and treated early. Flow cytometry is a key means of detecting AL, but requires manual processing by trained professionals, applying the manual gating method to huge amounts of raw data, making it time-consuming and labor-intensive. This study uses the deep learning Convolutional Neural Network (CNN) method for classification prediction based on two-dimensional graph for the parameter combinations of flow cytometry data to differentiate AL as normal, Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL) three categories. This approach normalizes raw data, removes outliers, and further augments data using Generative Adversarial Networks (GAN) to generate two-dimensional scattergrams of different parameter combinations generated by four methods. We then assess post-training classification prediction accuracy to identify the most effective preprocessing method and parameter combination. Experimental results obtain accuracy rates of 73–86 %. In the two-dimensional scatter plot of different parameter combinations generated by the four methods, the combination of CD3 and CD7 cell population parameters had the best classification and prediction accuracy, with image classification and identification reaching a high of 86 % through the use of GAN-based data augmentation. Advances in medical testing technology has reduced analysis times and increased data throughput. At the same time, artificial intelligence techniques are increasingly used for analysis and detection, performing large numbers of repetitive actions to reduce the potential for subjective assessment error from manual analysis, thus allowing for the earlier detection and treatment of disease.

Sistem Presensi Pendeteksi Wajah menggunakan Metode Modified Region Convolutional Neural Network dan PCA

Attendance is an activity to collect data to find out the number of attendances at an event or activity. In order to increase effectiveness and time efficiency is important, it is necessary to have an artificial intelligence-based presence by means of face detection. This study aims to create a face detection attendance system using modified region convolutional neural network (MR-CNN) and principal component analysis (PCA) methods. This type of research is development research by applying the concept of digital image based on face recognition. This research applies a method in deep learning, namely MR-CNN through camera media to take images and the extraction process using the PCA method to reduce image resolution. Images that include various individuals who want to be recognized are stored and then used as datasets. From the dataset, the MR-CNN model was trained with training data. Our findings are in the form of a web-based facial short-attendance system. Where are the calculation results in this system has an accuracy value of 96.1%, so it can be used well for facial identification, and can be used at SMK Bhakti Anindya.

Enhancing Coincidence Time Resolution of PET detectors using short-time Fourier transform and residual neural network

Currently, a key metric for assessing the performance of time-of-flight positron emission tomography (TOF-PET) detectors is the Coincidence Time Resolution (CTR). In this study, our aim is to enhance CTR in PET detectors through features in frequency domain. To achieve this goal, we introduce a novel approach that integrates short-time Fourier transform (STFT) with a residual neural network (ResNet) to extract features from digitized waveforms obtained from PET detectors. A comparative analysis is performed by comparing the proposed method with the conventional constant fraction discriminator (CFD) and a convolutional neural network (CNN) using real data measured from a pair of lutetium yttrium oxyorthosilicate (LYSO)-SiPM units. The proposed method achieves an average CTR of 187 ps, which is 30.4% better than the CFD method and 4.5% better than the CNN. The method also reduces the average bias by 17.9% compared to the CNN method. A Class Activation Mapping (CAM) is also introduced to analyze the importance of different frequency components in Time of Flight (TOF) estimation and the attention of network on each time point, which may not only underscore the utility of STFT in accurate timing for waveforms with different rise time but also provide insights into why CNN may outperform traditional timing methods. The influence of loss function and data organization are also studied. The enhancement of proposed method in CTR of PET will contribute to the improvement of Signal-to-Noise Ratio (SNR) of PET images.

Comparative analysis of convolutional neural network and DenseNet121 transfer learning in agriculture focusing on crop leaf disease identification

PurposePlant cultivation holds a pivotal role in agriculture, necessitating precise disease identification for the overall health of plants. This research conducts a comprehensive comparative analysis between two prominent deep learning algorithms, convolutional neural network (CNN) and DenseNet121, with the goal of enhancing disease identification in tomato plant leaves.Design/methodology/approachThe dataset employed in this investigation is a fusion of primary data and publicly available data, covering 13 distinct disease labels and a total of 18,815 images for model training. The data pre-processing workflow prioritized activities such as normalizing pixel dimensions, implementing data augmentation and achieving dataset balance, which were subsequently followed by the modeling and testing phases.FindingsExperimental findings elucidated the superior performance of the DenseNet121 model over the CNN model in disease classification on tomato leaves. The DenseNet121 model attained a training accuracy of 98.27%, a validation accuracy of 87.47% and average recall, precision and F1-score metrics of 87, 88 and 87%, respectively. The ultimate aim was to implement the optimal classifier for a mobile application, namely Tanamin.id, and, therefore, DenseNet121 was the preferred choice.Originality/valueThe integration of private and public data significantly contributes to determining the optimal method. The CNN method achieves a training accuracy of 90.41% and a validation accuracy of 83.33%, whereas the DenseNet121 method excels with a training accuracy of 98.27% and a validation accuracy of 87.47%. The DenseNet121 architecture, comprising 121 layers, a global average pooling (GAP) layer and a dropout layer, showcases its effectiveness. Leveraging categorical_crossentropy as the loss function and utilizing the stochastic gradien descent (SGD) Optimizer with a learning rate of 0.001 guides the course of the training process. The experimental results unequivocally demonstrate the superior performance of DenseNet121 over CNN.

Modelling of improved LSTM +1D convolution neural network methods for the diagnosis of SKF bearings

Modelling of improved LSTM +1D convolution neural network methods for the diagnosis of SKF bearings

Internet street view image fusion method using convolutional neural network

The use of image fusion technology in the area of information processing is continuing to advance in depth thanks to ongoing hardware advancements and related research. An enhanced convolutional neural network approach is developed to fuse visible and infrared images, and image pre-processing is carried out utilising an image alignment method with edge detection in order to gain more accurate and trustworthy image information. The performance of the fast wavelet decomposition, convolutional neural network, and modified convolutional neural network techniques is compared and examined using four objective assessment criteria. The experimental findings demonstrated that the picture alignment was successful with an offset error of fewer than 3 pixels in the horizontal direction and an angle error of less than 0.3∘ in both directions. The revised convolutional neural network method increased the information entropy, mean gradient, standard deviation, and edge detection information by an average of 46.13%, 39.40%, 19.91%, and 3.72%. The runtime of the modified approach was lowered by 19.42% when compared to the convolutional neural network method, which enhanced the algorithm’s performance and boosted the effectiveness of picture fusion. The image fusion accuracy reached 98.61%, indicating that the method has better fusion performance and is of practical value for improving image fusion quality.

Modelling a self-defined CNN for effectual classification of PCOS from ultrasound images.

Polycystic Ovary Syndrome (PCOS) is a medical condition that causes hormonal disorders in women in their childbearing years. The hormonal imbalance leads to a delayed or even absent menstrual cycle. Women with PCOS mainly suffer from extreme weight gain, facial hair growth, acne, hair loss, skin darkening, and irregular periods, leading to infertility in rare cases. Doctors usually examine ultrasound images and conclude the affected ovary but are incapable of deciding whether it is a normal cyst, PCOS, or cancer cyst manually. To have access to the high-risk crucial PCOS and to detect the condition and the treatment aimed at mitigating health hazards such as endometrial hyperplasia/cancer, infertility, pregnancy complications, and the long-term burden of chronic diseases such as cardiometabolic disorders linked with PCOS. The proposed Self-Defined Convolution Neural Network method (SD_CNN) is used to extract the features and machine learning models such as SVM, Random Forest, and Logistic Regression are used to classify PCOS images. The parameter tuning is done with lesser parameters in order to overcome over-fitting issues. The self-defined model predicts the occurrence of the cyst based on the analyzed features and classifies the class labels effectively. The Random Forest Classifier was found to be the most reliable and accurate among Support Vector Machine (SVM) and Logistic Regression (LR), with accuracy being 96.43%. The proposed model establishes better trade-off compared to various other approaches and works effectually for PCOS prediction.

High-Resolution Recognition of Orbital Angular Momentum Modes in Asymmetric Bessel Beams Assisted by Deep Learning

Abstract Fractional orbital angular momentum (OAM) vortex beams present a promising way to increase the data throughput in optical communication systems. Nevertheless, high-precision recognition of fractional OAM with different propagation distances remains a significant challenge. We develop a convolutional neural network (CNN) method to realize high-resolution recognition of OAM modalities, leveraging asymmetric Bessel beams imbued with fractional OAM. Experimental results prove that our method achieves a recognition accuracy exceeding 94.3% for OAM modes, with an interval of 0.05, and maintains a high recognition accuracy above 92% across varying propagation distances. The findings of our research will be poised to significantly contribute to the deployment of fractional OAM beams within the domain of optical communications.

Acoustic estimation of the manatee population and classification of call categories using artificial intelligence

The population sizes of manatees in many regions remain largely unknown, primarily due to the challenging nature of conducting visual counts in turbid and inaccessible aquatic environments. Passive acoustic monitoring has shown promise for monitoring manatees in the wild. In this study, we present an innovative approach that leverages a convolutional neural network (CNN) for the detection, isolation and classification of manatee vocalizations from long-term audio recordings. To improve the effectiveness of manatee call detection and classification, the CNN works in two phases. First, a long-term audio recording is divided into smaller windows of 0.5 seconds and a binary decision is made as to whether or not it contains a manatee call. Subsequently, these vocalizations are classified into distinct vocal classes (4 categories), allowing for the separation and analysis of signature calls (squeaks). Signature calls are further subjected to clustering techniques to distinguish the recorded individuals and estimate the population size. The CNN was trained and validated using audio recordings from three different zoological facilities with varying numbers of manatees. Three different clustering methods (community detection with two different classifiers and HDBSCAN) were tested for their suitability. The results demonstrate the ability of the CNN to accurately detect manatee vocalizations and effectively classify the different call categories. In addition, our study demonstrates the feasibility of reliable population size estimation using HDBSCAN as clustering method. The integration of CNN and clustering methods offers a promising way to assess manatee populations in visually challenging and inaccessible regions using autonomous acoustic recording devices. In addition, the ability to differentiate between call categories will allow for ongoing monitoring of important information such as stress, arousal, and calf presence, which will aid in the conservation and management of manatees in critical habitats.

Chaos prediction of motor based on the integrated method of convolutional neural network and multi-reservoir echo state network

Chaos prediction of motor based on the integrated method of convolutional neural network and multi-reservoir echo state network

Quantum-inspired vortex search algorithm with deep neural networks for multi-pose facial expression recognition

The rapid expansion of artificial intelligence technologies has enabled machines to comprehend emotional intelligence. Among various indicators, facial expressions serve as an effective medium for understanding emotions. The concept of facial expression recognition (FER) relies heavily on the accurate and robust features available. Initially, the method of three-channel convolutional neural networks (TC-CNN) is adapted to extract facial features. However, only extracting the features is insufficient, the optimization of the extracted features is crucial to determining precise and robust features. This research work focuses on the optimization of the features using the quantum-inspired vortex search algorithm (QVSA). The QVSA integrates the attributes of Q-bits into the vortex search algorithm (VSA), optimizing the features by using the Q-bits to determine the vortex center on the Bloch sphere. The Q-bit attributes also improve the diversity of the features and help to avoid the premature convergence of the VSA. The final recognition of the facial expressions is performed using the deep neural network method of ResNet101v2. The experiments for facial expression recognition are performed on the datasets of RaFD and KDEF, which include different facial positions such as front pose, diagonal pose and profile pose. Performance comparisons demonstrate the effectiveness of the proposed system over state-of-the-art facial expression techniques.