Convolution Algorithm Research Articles

Improved faster region convolutional neural network algorithm for UAV target detection in complex environment

With the development of artificial intelligence technology and Unmanned Aerial Vehicle (UAV) technology, the traditional UAV target detection methods are difficult to achieve high detection accuracy in complex and changeable scenes. The paper selects Faster Region Convolutional Neural Network (Faster R-CNN) as the basic algorithm due to its high scalability and excellent classification performance. The convolution mode of the convolution layer in the network structure is adjusted to deformable convolution. Additionally, a hybrid attention mechanism module is added to combine channel attention and spatial attention modules behind the output layer of the Faster R-CNN network. Finally, the Soft Non-Maximum Suppression (Soft-NMS) method is selected to combine the improved Faster R-CNN algorithm of multiple individuals into the final target detection model. The performance of the improved Faster R-CNN algorithm and the original Faster R-CNN algorithm was verified through the VisDrone 2018-DET dataset and the full class average accuracy Mean Average Precision (mAP), accuracy and precision. The accuracy and logarithm of loss values of the improved Faster R-CNN algorithm and the original Faster R-CNN algorithm's Region Proposal Network (RPN) were 0.985, 0.981, 0.018, and 0.052, respectively. The target detection model based on the hybrid attention mechanism of ResNet-50 network fusion Spartial Attention Module (SAM) and Channel Attention Module (CAM) had the best classification performance, with the accuracy, precision and overall average accuracy of 25.8 %, 24.6 % and 22.7 %, respectively. The results are helpful to improve the target detection ability of UAV in complex environment, and contribute to the development of target detection technology in the future.

Implementation of Convolutional Neural Network Algorithm for Detecting Empty Parking Area Based on Raspberry Pi

Implementation of Convolutional Neural Network Algorithm for Detecting Empty Parking Area Based on Raspberry Pi

Adaptive Graph Convolution Algorithm Based on 3D Vision Selectivity and Its Application in Scene Segmentation

Adaptive Graph Convolution Algorithm Based on 3D Vision Selectivity and Its Application in Scene Segmentation

Efficient Velvet-Noise Convolution in Multicore Processors

Velvet noise, a sparse pseudo-random signal, finds valuable applications in audio engineering, such as artificial reverberation, decorrelation filtering, and sound synthesis. These applications rely on convolution operations whose computational requirements depend on the length, sparsity, and bit resolution of the velvet-noise sequence used as filter coefficients. Given the inherent sparsity of velvet noise and its occasional restriction to a few distinct values, significant computational savings can be achieved by designing convolution algorithms that exploit these unique properties. This paper shows that an algorithm called the transposed double-vector filter is the most efficient way of convolving velvet noise with an audio signal. This method optimizes access patterns to take advantage of the processor’s fast caches. The sequential sparse algorithm is shown to be always faster than the dense one, and the speedup is linearly dependent on sparsity. The paper also explores the potential for further speedup on multicore platforms through parallelism and evaluate the impact of data encoding, including 16-bit and 32-bit integers and 32-bit floating-point representations. The results show that using the fastest implementation of a long velvet-noise filter, it is possible to process more than 40 channels of audio in real time using the quad-core processor of a modern system-on-chip.

Identification of Regulatory RNA-Binding Proteins Associated with Immune Infiltration in Laryngeal Squamous Cell Carcinoma.

We analyzed bulk RNA sequencing and single-cell RNA sequencing (scRNA-seq) data to identify alternative splicing (AS) events and regulatory RNA-binding proteins (RBPs) associated with immune infiltration in human laryngeal squamous cell carcinoma (LSCC). Whole-transcriptome sequencing data of 20 human laryngeal cancer and paracancerous tissues were downloaded from the Gene Expression Omnibus public database, using newly published splicing-site usage variation analysis software to obtain highly conserved regulated AS (RAS) events, and scientific reverse convolution algorithm analysis was used to identify significantly different immune cells and perform a correlation analysis between the two. The software package edgeR was used to identify differentially expressed RBPs and the immune infiltration-related LSCC-RAS they may regulate. Finally, we present the expression profiles and survival curves of 117 human laryngeal cancer samples from The Cancer Genome Atlas dataset for the identified RBPs and LSCC-RAS. We also downloaded the gene set enrichment 150321 scRNA-seq data for two human LSCC tissue samples. The RBP expression pattern and the expression of prophase RBP genes were analyzed in different LSCC cell populations. RNA-binding motif protein 47 (RBM47) and filamin A, as well as the RBP-RAS events that were screened in both the fibulin 2 and fibronectin 1 genes, were all significantly associated with the prognosis, and the RBM47 gene was upregulated in myeloid cells. Because the prognosis was significantly associated with two RBP regulators and two LSCC-RAS events, they may be critical regulators of immune cell survival during laryngeal cancer progression, and RBM47 may regulate macrophage-associated AS and affect immunity.

Programming knowledge tracing based on heterogeneous graph representation

Existing programming knowledge tracing methods have not deeply explored the relationships between knowledge concepts and programming questions or students’ codes, which leads to insufficient representation of programming questions and students’ codes and insufficient tracing of students’ knowledge states. To address these problems, this study proposes a programming knowledge tracing method based on a heterogeneous graph representation. We constructed a heterogeneous graph, where nodes represent the knowledge concept, programming question, and student code, and edges represent the relationships among knowledge concepts, programming questions, and student code. Subsequently, we leverage a graph structure learning algorithm to exploit the potential connections among them, aiming to enhance the graph structure information. Based on the graph, we further fully integrated the knowledge concept information into the representations of programming questions and student codes through the graph convolutional representation learning algorithm. Moreover, an attention mechanism is utilized to compute the effect of programming questions and student codes on the knowledge state and update the student's knowledge state in conjunction with the corresponding answer results. Furthermore, the proposed method provides an assessment of the mastery level for each knowledge concept rather than only assessing each question using traditional methods. Experiments on two datasets, CodeWorkout and CodeNet, show that the proposed method achieves an average improvement of 4.28% in ACC and 6.48% in AUC compared with the baseline models, validating the effectiveness of the proposed method.

Classification of Marine Fish Based on Image Using Convolutional Neural Network Algorithm and VGG16

Marine fish are one of the many natural resources that frequently consumed by people. However, several types of marine fish are prohibited from consumption due to being nearly extinct. Additionally, some fish species contain high levels of mercury that can be harmful to human if consumed. Because of the large number of marine fish species, it becomes challenging to identify them without knowledge of fisheries. Moreover, computers have become highly advanced devices that facilitate various human activities. This advancement allows for the creation of systems capable of processing information from images, known as image classification. There are numerous methods that can be employed in designing an image classification system, one of which is transfer learning. This study was aimed to design an image classification system using the transfer learning method with a pre-trained VGG16 model and Convolutional Neural Network algorithm. The research results showed a training data accuracy of 100% and a validation data accuracy of 99.3%, with an overall accuracy of 84% and a loss value of 0.6591.

Application of Fashion Element Trend Prediction Model Integrating AM and EfficientNet-b7 Models in Art Design

The poor application effect of traditional sports training methods and the difficulty of recording data due to the time and space constraints of sports make it difficult for trainers to improve their learning outcomes. Based on this, the study proposes to apply human posture recognition in sports teaching design, and use VGGNet-19 as a feature extractor and OpenCV open source software to capture posture movements, and introduce the concepts of joint angle and movement similarity to design a sports assessment system for physical education based on the geometric spatial feature variability analysis of posture based on limb angle information. The testing outcomes demonstrate that the study’s improved gesture recognition algorithm has a recognition rate of more than 90% on gesture movements, and the maximum recognition error value (0.010) is smaller than that of the dynamic time-regularised gesture algorithm (0.014) and the convolutional neural network algorithm (0.017). The assessment system is also better able to improve students’ professional performance and satisfaction, with its average professional score and satisfaction reaching 86 and 92%, which is significantly better than other comparative algorithms. The method is effective in providing trainers with data-based training scenarios and helping them to improve their learning in sport.

Inferring single-cell spatial gene expression with tissue morphology via explainable deep learning.

The spatial arrangement of cells is vital in developmental processes and organogenesis in multicellular life forms. Deep learning models trained with spatial omics data uncover complex patterns and relationships among cells, genes, and proteins in a high-dimensional space, providing new insights into biological processes and diseases. State-of-the-art in silico spatial multi-cell gene expression methods using histological images of tissue stained with hematoxylin and eosin (H&E) to characterize cellular heterogeneity. These computational techniques offer the advantage of analyzing vast amounts of spatial data in a scalable and automated manner, thereby accelerating scientific discovery and enabling more precise medical diagnostics and treatments. In this work, we developed a vision transformer (ViT) framework to map histological signatures to spatial single-cell transcriptomic signatures, named SPiRiT ( S patial Omics P rediction and R eproducibility integrated T ransformer). Our framework was enhanced by integrating cross validation with model interpretation during hyper-parameter tuning. SPiRiT predicts single-cell spatial gene expression using the matched histopathological image tiles of human breast cancer and whole mouse pup, evaluated by Xenium (10x Genomics) datasets. Furthermore, ViT model interpretation reveals the high-resolution, high attention area (HAR) that the ViT model uses to predict the gene expression, including marker genes for invasive cancer cells ( FASN ), stromal cells ( POSTN ), and lymphocytes ( IL7R ). In an apple-to-apple comparison with the ST-Net Convolutional Neural Network algorithm, SPiRiT improved predictive accuracy by 40% using human breast cancer Visium (10x Genomics) dataset. Cancer biomarker gene prediction and expression level are highly consistent with the tumor region annotation. In summary, our work highlights the feasibility to infer spatial single-cell gene expression using tissue morphology in multiple-species, i.e., human and mouse, and multi-organs, i.e., mouse whole body morphology. Importantly, incorporating model interpretation and vision transformer is expected to serve as a general-purpose framework for spatial transcriptomics.

Healthcare Analysis Based on Diabetes Prediction Using a Cuckoo-Based Deep Convolutional Long-Term Memory Algorithm

In recent years, the demand for mobile medical applications utilizing the Internet of Things (IoT) for diabetes diagnosis has been progressively increasing. Diabetes is commonly known as a chronic illness that presents a significant danger to individuals, analysing when blood sugar levels surpass the typical range. If diabetes is not promptly treated, hyperosmolar conditions can lead to serious health issues like hyperglycaemia and possibly even death. Since early detection enables lifestyle changes that prevent the disease's progression, it is crucial for diabetes management and health systems. However, diabetes diagnosis has a long computational time and low prediction accuracy. To address this issue, we propose a Cuckoo-Based Deep Convolutional Long-Term Memory (CDC-LSTM) algorithm that increases accuracy by classifying diabetics or non-diabetics. Additionally, we utilize the Standardized Feature Scaler (SFS) method to normalize the variance data by removing the mean of each feature. Moreover, we select the optimal features in the diabetes dataset utilising the Filter-Based Decision Tree (FBDT) technique. Finally, the proposed CDC-LSTM method can be used to distinguish between diabetics and non-diabetics, improving the accuracy of identifying diabetic patients. Additionally, the proposed method can predict diabetes using performance assessments such as precision, recall, and F-measure. Furthermore, the method's accuracy can be improved to 95.18% compared to previous approaches.

Sleeping Stage Classification Using CNN

Abstract: Maintaining proper health and mental stability is critical for overall health and well-being. Despite a good deal of research investment, sleep quality continues to be a crucial public challenge. Nowadays, people of all age groups are affected by improper sleep quality. Poor sleep can lead to a variety of neurological disorders. Sleep disorders are common in all subsets of the population, independently of gender. This public health challenge greatly affects quality of life in terms of both physical and mental health. Insomnia, parasomnias, sleep- related breathing difficulties, hypersomnia, bruxism, narcolepsy, and circadian rhythm disorders are some common examples of sleep-related disorders. Some of these disorders can be treated with proper analysis of early symptoms; in such cases, adequate sleep quality is essential for the patient’s recovery. Artificial intelligence has several applications in sleep medicine including sleep and respiratory event scoring in the sleep laboratory, diagnosing and managing sleepdisorders, and population health. While still in its nascent stage, there are several challenges which preclude AI’s generalizability and wide-reaching clinical applications. Artificial intelligence is a powerful tool in healthcare that may improve patient care, enhance diagnostic abilities, and augment the management of sleep disorders. However, there is a need to regulate and standardize existing machine learning algorithms and deep learning algorithm prior to its inclusion in the sleep clinic. In this project, we can develop the framework for sleeping stage classification for both subjective data and ECG data and classify the data using Convolutional neural network algorithm to analyse the multiple sleeping stages.

An efficient full-size convolutional computing method based on memristor crossbar

Modern artificial intelligence systems based on neural networks need to perform a large number of repeated parallel operations quickly. Without hardware acceleration, they cannot achieve effectiveness and availability. Memristor-based neuromorphic computing systems are one of the promising hardware acceleration strategies. In this paper, we propose a full-size convolution algorithm (FSCA) for the memristor crossbar, which can store both the input matrix and the convolution kernel and map the convolution kernel to the entire input matrix in a full parallel method during the computation. This method dramatically increases the convolutional kernel computations in a single operation, and the number of operations no longer increases with the input matrix size. Then a bidirectional pulse control switch integrated with two extra memristors into CMOS devices is designed to effectively suppress the leakage current problem in the row and column directions of the existing memristor crossbar. The spice circuit simulation system is built to verify that the design convolutional computation algorithm can extract the feature map of the entire input matrix after only a few operations in the memristor crossbar-based computational circuit. System-level simulations based on the MNIST classification task verify that the designed algorithm and circuit can effectively implement Gabor filtering, allowing the multilayer neural network to improve the classification task recognition accuracy to 98.25% with a 26.2% reduction in network parameters. In comparison, the network can even effectively immunize various non-idealities of the memristive synaptic within 30%.

Bioinspired composite fiber aerogel pressure sensor for machine-learning-assisted human activity and gesture recognition

Machine-learning-assisted human activity and gesture recognition are valuable for human-computer interaction, and data acquisition often necessitates high-performance sensors. Here, inspired by spider web and bat wing airflow sensing system, polyimide fiber (PIF)/carbon black (CB) composite fiber aerogel (CFA) pressure sensor with biomimetic hair-Merkel cell sensitive unit was developed, exhibiting ultralow detection limit (2 Pa), high pressure sensitivity (S=23.1 kPa-1), wide linear detection capacity up to 67.61 kPa, and fast response/recovery time (140/100 ms). Thanks to the excellent mechanical property and environmental tolerance of CFA, it also possesses excellent low fatigue over >4000 cycles and good durability even at extreme high-temperature (200 °C) and underwater conditions. The superior signal data of the sensor, combined with the Convolutional Neural Network machine learning algorithm, achieves ultra-high prediction accuracies of 96.73% and 98.26% for human activity and gesture recognition, respectively. Additionally, CFA also has amazing thermal management properties, making it to be an ideal candidate for wearable electronics with excellent wearing comfort and safety.

An Efficient Algorithm for Acute Lymphoblastic Leukemia Quantification in Blood Cell Images Using Support Vector Machines Algorithm

As the capabilities of information technologies develop, the effectiveness of using the capabilities of CNN convolutional neural networks, an algorithm with a high level of accuracy, and early prediction of various diseases in blood cell images is high. This study proposes an effective method for early detection of acute lymphoblastic leukemia in blood cell images using Support Vector Machines (SVM) algorithm. In this method, a pattern of cells is recognized and used to identify cell markers specific to leukemia. This algorithm is used to match leukemia to a single marker in the cell image. Comparing SNN convolutional neural network algorithms with random forest (RF), Bayesian classifier, Support Vector Machines (SVM) and K nearest neighbor (KNN) algorithms, the results obtained by Support Vector Machines (SVM) were found to be 90.9% efficient.

Application of an Improved Convolutional Neural Network Algorithm in Text Classification

This paper proposes a text classification model based on a combination of a convolutional neural network (CNN) and a support vector machine (SVM) using Amazon review polarity, TREC, and Kaggle as experimental data. By adding an attention mechanism to simplify the parameters and using the classifier based on SVM to replace the Softmax layer, the extraction effect of feature words is improved and the problem of weak generalization ability of the CNN model is solved. Simulation experiments show that the proposed algorithm performs better in precision rate, recall rate, F1 value, and training time compared with CNN, RNN, BERT and term frequency-inverse document frequency (TF-IDF).

Parallel GEMM-based convolutions for deep learning on multicore ARM and RISC-V architectures

We present high performance, multi-threaded implementations of three GEMM-based convolution algorithms for multicore processors with ARM and RISC-V architectures. The codes are integrated into CONVLIB, a library that has the following unique features: (1) scripts to automatically generate a key component of GEMM, known as the micro-kernel, which is typically written in assembly language; (2) a modified analytical model to automatically tune the algorithms to the underlying cache architecture; (3) the ability to select four hyper-parameters: micro-kernel, cache parameters, parallel loop, and GEMM algorithm dynamically between calls to the library, without recompiling it; and (4) a driver to identify the best hyper-parameters. In addition, we provide a detailed performance evaluation of the convolution algorithms, on five ARM and RISC-V processors, and we publicly release the codes.

Criminal Identification Web App Utilizes Facial Recognition to Identify and Track Criminals

Criminal record generally contains all the information both personal and criminal with the photograph of the person. In order to recognize Criminal, identification of some sort is required, designated by eyewitnesses. In most cases the resolution or/and quality of the recorded image sections is unsatisfactory and is difficult to recognize the face. Recognition can be achieved in various different ways like DNA, eyes, finger print, etc. One of the ways is face identification. Since facial recognition technology is powered by artificial intelligence, it can provide excellent results in identifying criminals. Even considering that most people, when committing an illicit activity, try to hide their identity: hiding their faces or covering their faces with scarves, masks, etc. In such cases, AI uses deep learning methods to identify the individual. In this project, proposed a CrimeNet an automatic criminal identification system for Police Department to enhance and upgrade the criminal classification into a more effective and efficient approach using Convolutional neural network algorithms.

Pneumonia Detection Using Deep Learning

Lung disease called pneumonia is brought on by a bacterial infection. An effective treatment plan heavily depends on an early diagnosis. An expert radiologists can typically diagnose the condition with a chest x-ray. For several reasons, including the emergence of an illness that is not visible on a chest x-ray or the possibility that it is mistaken for another ailment, the diagnosis may be arbitrary. Thus, to assist clinicians, automated diagnostic technologies are required. In this work, a deep learning system for picture classification the Convolutional Neural Networks algorithm was constructed on a dataset. Chest radiographs are included in the data set, and the created model is assessed using a few statistical factors. Key Words: Diagnosis, computer-aided, deep learning, pneumonia, convolution neural networks, mass chest x-ray, chest x-ray14.

A Self‐Powered Dual Ratchet Angle Sensing System for Digital Twins and Smart Healthcare

AbstractIn the swiftly progressing landscape of wearable electronics and the Internet of Things (IoTs), there is a burgeoning demand for devices that are lightweight, cost‐effective, and self‐powered. In this study, a self‐powered bidirectional knee joint motion monitoring system is introduced, leveraging a dual ratchet sensing (DRS) system fabricated using 3D printing technology. This approach offers substantial economic and portability benefits. The DRS system is engineered to harness the negative work generated from knee joint movements to power commercial electronic devices, obviating the need for additional metabolic energy from the human body. By synergizing the DRS with virtual reality technology, it becomes feasible to monitor knee joint movements in real‐time with remarkable accuracy, presenting a novel avenue for the integration of digital twin technology. Through the amalgamation of convolutional neural network machine learning algorithms with Bayesian optimization techniques, the DRS system can discern up to 97% of knee joint movements, paving the way for innovative applications in smart rehabilitation and healthcare.

Hydrogel-based flexible degradable triboelectric nanogenerators for human activity recognition

The development of Internet of Things (IoT) and Artificial Intelligence (AI) technologies has led to an increased demand for wearable electronic devices that can be used in a variety of contexts. Hydrogel sensors have been regarded as an ideal material for the fabrication of multifunctional flexible wearable devices. However, it is difficult to create hydrogel sensors with ideal electrical and mechanical properties. In this study, a conductive hydrogel prepared by a simple method has been introduced. The hydrogel was immersed in a binary solvent of glycerol, water and sodium citrate to form a transparent, long-term stable and highly conductive gelatin-NaCl organohydrogel (GNOH). A stretchable and durable gelatin Ecoflex triboelectric nanogenerator (GE-TENG) was created by combining an organic hydrogel with Ecoflex elastomer. It has high electrical and mechanical performance, with an open-circuit voltage of 142 V, current of 3.8 μA, power output of 19.36 μW and response time of 85 ms. Moreover, the single-electrode mode of the GE-TENG provides high electrical output for powering portable electronic devices and can capture biomechanical energy in temperatures as low as −20 °C. As a result of these capabilities, GE-TENG can effectively identify human movements with precision and transmit signals wirelessly, thereby broadening its potential applications. The GE-TENG-based self-powered intelligent sensing system (GSIS) for human-computer interaction demonstrates a wearable intelligent system for controlling the movement of a tank model and a manipulator by integrating the signal acquisition, signal processing and signal output components. The system achieves 100% signal recognition accuracy when combined with covariance matrix feature analysis and Convolutional Neural Network classification algorithms. This research demonstrates the potential of wearable hydrogel-based electronic devices for monitoring human motion, harvesting biomechanical energy and human-computer interaction.