Neural Network Model Research Articles

Artificial intelligence methods available for cancer research.

Cancer is a heterogeneous and multifaceted disease with a significant global footprint. Despite substantial technological advancements for battling cancer, early diagnosis and selection of effective treatment remains a challenge. With the convenience of large-scale datasets including multiple levels of data, new bioinformatic tools are needed to transform this wealth of information into clinically useful decision-support tools. In this field, artificial intelligence (AI) technologies with their highly diverse applications are rapidly gaining ground. Machine learning methods, such as Bayesian networks, support vector machines, decision trees, random forests, gradient boosting, and K-nearest neighbors, including neural network models like deep learning, have proven valuable in predictive, prognostic, and diagnostic studies. Researchers have recently employed large language models to tackle new dimensions of problems. However, leveraging the opportunity to utilize AI in clinical settings will require surpassing significant obstacles-a major issue is the lack of use of the available reporting guidelines obstructing the reproducibility of published studies. In this review, we discuss the applications of AI methods and explore their benefits and limitations. We summarize the available guidelines for AI in healthcare and highlight the potential role and impact of AI models on future directions in cancer research.

Exploring the accuracy of tooth loss prediction between a clinical periodontal prognostic system and a machine learning prognostic model.

The aim of this analysis was to compare a clinical periodontal prognostic system and a developed and externally validated artificial intelligence (AI)-based model for the prediction of tooth loss in periodontitis patients under supportive periodontal care (SPC) for 10 years. Clinical and radiographic parameters were analysed to assign tooth prognosis with a tooth prognostic system (TPS) by two calibrated examiners from different clinical centres (London and Pittsburgh). The prediction model was developed on the London dataset. A logistic regression model (LR) and a neural network model (NN) were developed to analyse the data. These models were externally validated on the Pittsburgh dataset. The primary outcome was 10-year tooth loss in teeth assigned with 'unfavourable' prognosis. A total of 1626 teeth in 69 patients were included in the London cohort (development cohort), while 2792 teeth in 116 patients were included in the Pittsburgh cohort (external validated dataset). While the TPS in the validation cohort exhibited high specificity (99.96%), moderate positive predictive value (PPV = 50.0%) and very low sensitivity (0.85%), the AI-based model showed moderate specificity (NN = 52.26%, LR = 67.59%), high sensitivity (NN = 98.29%, LR = 91.45%), and high PPV (NN = 89.1%, LR = 88.6%). AI-based models showed comparable results with the clinical prediction model, with a better performance in specific prognostic risk categories, confirming AI prediction model as a promising tool for the prediction of tooth loss.

Multimodal urban mobility solutions for a smart campus using artificial neural networks for route determination and an algorithm for arrival time prediction

Artificial intelligence algorithms play a key role in solving multimodal urban mobility problems, this study outlines a scheme covering bus, boat, and pedestrian transport modes. This involves designing an Artificial Neural Network (ANN) model to classify the most suitable routes for passengers, together with an algorithm capable of estimating the arrival times of various transportation modes at designated stopping points or buildings within the smart campus. The information used to train the ANN is obtained from an Internet of Things (IoT) network and a database that includes the available routes within the campus (which is situated in the Brazilian Amazon), where there is a multimodal electric mobility service. In seeking to achieve the objectives, this ANN relied on the user's geographic location in the input layer, and route mapping data stored in the database in the output layer, as well as the backpropagation algorithm for computational processing and adjustments of synaptic weight. The algorithm for modal arrival time prediction used real-time geographic location data, together with the Haversine formula to calculate geographic distance, while the average speed was obtained from the IoT network. A range of different ANN parameters were included in the experiments. In the study case, the ANN results show an improvement of above 92% for determining the best routes and predicting arrival times. The findings show that ANNs can effectively find the best route within a real-world smart campus environment. Moreover, the results show the arrival time can be estimated by using real-time geographic location data.

Enhancing Asian Indigenous Language Processing through Deep Learning-based Handwriting Recognition and Optimization Techniques

Asian indigenous language or autochthonous language is a language which is native to a region and spoken by indigenous people in Asia. This language is a linguistically different community created in the region. Recently, researchers in handwriting detection studies comparing with indigenous languages have attained important internet amongst the research community. A new development of artificial intelligence (AI), natural language processing (NLP), cognitive analytics, and computational linguistics (CL) find it helpful in the analysis of regional low-resource languages. It can be obvious in the obtainability of effectual machine detection methods and open access handwritten databases. Tamil is the most ancient Indian language that is mostly exploited in the Southern part of India, Sri Lanka, and Malaysia. Tamil handwritten Character Recognition (HCR) is a critical procedure in optical character detection. Therefore, this study designs a Henry Gas Solubility Optimization with Deep Learning-based Handwriting Recognition Model (HGSODL-HRM) for Asian Indigenous Language Processing. The proposed HGSODL-HRM technique relies on computer vision and DL concepts for automated handwriting recognition in the Tamil language, which is one of the popular indigenous languages in Asia. To accomplish this, the HGSODL-HRM technique employs a capsule network (CapsNet) model for feature vector generation with the HGSO algorithm as a hyperparameter optimizer. For the recognition of handwritten characters, wavelet neural network (WNN) model is exploited. Finally, the WNN parameters can be optimally chosen by sail fish optimizer (SFO) algorithm. To demonstrate the promising results of the HGSODL-HRM system, an extensive range of simulations can be implemented. The simulation outcomes stated the betterment of the HGSODL-HRM system compared to recent DL models.

Architecture Optimization for Hybrid Deep Residual Networks in Liver Tumor Segmentation Using a GA

Nowadays, liver tumor segmentation has been widely applied for vital medical objectives such as illness diagnosis, treatment, and evaluation of liver function. In this work, we aim to improve the performance of hybrid convolution neural network (CNN) models for liver tumor segmentation. A new automatic design method is introduced to build a hybrid CNN model with an optimal architecture using a combination of the original U_Net and the Res-UNet. In the proposed design method, the Res-UNet model is divided into different blocks, and then many different candidate architectures for the hybrid CNN model can be created using gradually these blocks to replace their corresponding blocks of the original U-Net. The replacement is executed if the candidate architecture using a residual block gives better performance than the corresponding one that uses the original U-Net block. We use a genetic algorithm (GA) to get the optimal architecture among the candidate architectures that gives the best performance for the hybrid CNN model. During the optimization process, in addition to the optimal architecture, other crucial configurations such as the learning algorithm, learning rate, and batch size are also determined. We refer to the hybrid CNN model that has the best architecture as GA_Res_UNet. Comparison has been made with other CNN models that are commonly used in image segmentation of liver tumors. We use the 3D-IRCADb01 liver tumor dataset that contains 2800 magnetic resonance (MR) images of 20 patients. Each image has a liver tumor with its corresponding mask. All MR images are resized to 256 × 256 in DICOM format. The findings indicate that GA_Res_UNet outperformed the other compared models over the considered dataset. It achieves 98.5% of dice coefficient and 99.8% of accuracy. Moreover, two other liver tumor segmentation datasets (LiTS17 and CHAOS) are utilized in another comparison to show the superior performance of our proposed model. The reported results show that the proposed model outperformed the other compared state-of-the-art models.

A performance neural network model for conventional solar stills via transfer learning

Predictive solar-desalination models are becoming more widespread using ML and AI. However, forecasting solar still water productivity based on numerous designs still needs to be improved. Herein, we used transfer learning to create precise supervised predictive ANN regression models for water productivity (L/m2.day) predictions based on literature findings. Such observation datasets from single-basin solar stills were utilized to build the random initialization ANN model. The transfer learning method was applied to the latter model by taking the learned network (weights) for fine-tuning the hyperparameters from the earlier developed novel hybrid solar still known as the source (pre-trained) ANN model, to predict the target ANN model. Based on most minor statistical errors, the pre-trained model with 5–64–64-1 architecture and ReLU activation function was the most appropriate for water productivity prediction. All created ANN models were compared to the MLR model. The results revealed that the generated target ANN model outperformed the ANN RI and MLR with OI values of 0.872, 0.834, and 0.803, respectively, in all modeling stages. The target ANN model's accuracy and generalization were sufficient. The target ANN model had residuals of forecasted distillate values of around 1%. This work discusses the significance of transfer learning to generate accurate target ANN models for predicting freshwater outputs in single-slope solar stills, which can be integrated with established theoretically tuned parameters to enhance performance and maximize distillate water yields.

Identification of Euphorbiae pekinensis Radix and its counterfeit and adulterated products based on DNA barcode, UPLC-Q-TOF-MS, UPLC fingerprint, and chemometrics.

Euphorbiae pekinensis Radix (EPR) is a traditional Chinese herb commonly used to treat edema, pleural effusion, and ascites. However, counterfeit and adulterated products often appear in the market because of the homonym phenomenon, similar appearance, and artificial forgery of Chinese herbs. This study comprehensively evaluated the quality of EPR using multiple methods. The DNA barcode technique was used to identify EPR, while the UPLC-Q-TOF-MS technique was utilized to analyze the chemical composition of EPR. A total of 15 tannin and phenolic acid components were identified. Furthermore, UPLC fingerprints of EPR and its common counterfeit products were established, and unsupervised and supervised pattern recognition models were developed using these fingerprints. The backpropagation artificial neural network and counter-propagation artificial neural network models accurately identified counterfeit and adulterated products, with a counterfeit ratio of more than 25%. Finally, the contents of the chemical markers 3,3'-di-O-methyl ellagic acid-4'-O-β-D-glucopyranoside, ellagic acid, 3,3'-di-O-methyl ellagic acid-4'-O-β-d-xylopyranoside, and 3,3'-di-O-methyl ellagic acid were determined to range from 0.05% to 0.11%, 1.95% to 8.52%, 0.27% to 0.86%, and 0.10% to 0.42%, respectively. This proposed strategy offers a general procedure for identifying Chinese herbs and distinguishing between counterfeit and adulterated products.

Exploring the effectiveness of reward-based learning strategies for second-language speech sounds.

Adults struggle to learn non-native speech categories in many experimental settings (Goto, Neuropsychologia, 9(3), 317-323 1971), but learn efficiently in a video game paradigm where non-native speech sounds have functional significance (Lim & Holt, Cognitive Science, 35(7), 1390-1405 2011). Behavioral and neural evidence from this and other paradigms point toward the involvement of reinforcement learning mechanisms in speech category learning (Harmon, Idemaru, & Kapatsinski, Cognition, 189, 76-88 2019; Lim, Fiez, & Holt, Proceedings of the National Academy of Sciences, 116, 201811992 2019). We formalize this hypothesis computationally and implement a deep reinforcement learning network to map between environmental input and actions. Comparing to a supervised model of learning, we show that the reinforcement network closely matches aspects of human behavior in two experiments - learning of synthesized auditory noise tokens and improvement in speech sound discrimination. Both models perform comparably and the similarity in the output of each model leads us to believe that there is little inherent computational benefit to a reward-based learning mechanism. We suggest that the specific neural circuitry engaged by the paradigm and links between striatum and superior temporal areas play a critical role in effective learning.

An Attention-Based Deep Learning Model for Phase-Resolved Wave Prediction

Abstract Phase-resolved wave prediction capability, even if only over two wave periods in advance, is of value for optimal control of wave energy converters, resulting in a dramatic increase in power generation efficiency. Previous studies on wave-by-wave predictions have shown that an artificial neural network (ANN) model can outperform the traditional linear wave theory-based model in terms of both prediction accuracy and prediction horizon when using synthetic wave data. However, the prediction performance of ANN models is significantly reduced by the varying wave conditions and buoy positions that occur in the field. To overcome these limitations, a novel wave prediction method is developed based on the neural network with an attention mechanism. This study validates the new model using wave data measured at sea. The model utilizes past time histories of three Sofar Spotter wave buoys at upwave locations to predict the vertical motion of a Datawell Waverider-4 at a downwave location. The results show that the attention-based neural network model is capable of capturing the slow variation in the displacement of the buoys, which reduces the prediction error compared to a standard ANN and long short-term memory model.

Predicting Particle Size and Soil Organic Carbon of Soil Profiles Using VIS-NIR-SWIR Hyperspectral Imaging and Machine Learning Models

Modeling spectral reflectance data using machine learning algorithms presents a promising approach for estimating soil attributes. Nevertheless, a comprehensive investigation of the most effective models, parameters, wavelengths, and data acquisition techniques is essential to ensure optimal predictive accuracy. This work aimed to (a) explore the potential of the soil spectral signature obtained in different spectral bands (VIS-NIR, SWIR, and VIS-NIR-SWIR) and, by using hyperspectral imaging and non-imaging sensors, in the predictive modeling of soil attributes; and (b) analyze the accuracy of different ML models in predicting particle size and soil organic carbon (SOC) applied to the spectral signature of different spectral bands. Six soil monoliths, located in the central north region of Parana, Brazil, were collected and scanned via hyperspectral cameras (VIS-NIR camera and SWIR camera) and spectroradiometer (VIS-NIR-SWIR) in the laboratory. The spectral signature of the soils was analyzed and subsequently applied to ML models to predict particle size and SOC. Each set of data obtained by the different sensors was evaluated separately. The algorithms used were k-nearest neighbors (KNN), support vector machine (SVM), random forest (RF), linear regression (LR), artificial neural network (NN), and partial least square regression (PLSR). The most promising predictive performance was observed for the complete VIS-NIR-SWIR spectrum, followed by SWIR and VIS-NIR. Meanwhile, KNN, RF, and NN models were the most promising algorithms in estimating soil attributes for the dataset obtained from both sensors. The general mean R2 (determination coefficient) values obtained using these models, considering the different spectral bands evaluated, were around 0.99, 0.98, and 0.97 for sand prediction, and around 0.99, 0.98, and 0.96 for clay prediction. The lower performances, obtained for the datasets from both sensors, were observed for silt and SOC, with R2 results between 0.40 and 0.59 for these models. KNN demonstrated the best predictive performance. Integrating effective ML models with robust sample databases, obtained by advanced hyperspectral imaging and spectroradiometers, can enhance the accuracy and efficiency of soil attribute prediction.

PREDIKSI KANDUNGAN MINYAK DAN ASAM LEMAK BEBAS BUAH SAWIT SECARA NON-DESTRUKTIF MENGGUNAKAN NEAR-INFRARED SPECTROSCOPY DAN METODE KALIBRASI HYBRID

The study used near-infrared reflectance spectroscopy and hybrid calibration methods to predict oil and free fatty acid content of oil palm fruitlets non-destructively. The reflectance and chemical content of oil palm fruitlets were measured and the calibration between near infrared spectra and chemical content was performed using hybrid calibration methods. The developed hybrid calibration model was validated to predict oil and free fatty acid of other oil palm fruitlets. The best models to predict oil and free fatty acid of oil palm fruitlets respectively were the hybrid model of Partial Least Square-Artificial Neural Network with 25 factor components (R2 = 0.96; SEP= 2.21%, RPD = 4.79) and 19 factor components (R2 = 0.96; SEP= 0.25%, RPD = 4.24) using Savitzky-Golay first derivative spectra pre-treatment.

Optimized Radio Frequency Footprint Identification Based on UAV Telemetry Radios

With the widespread use of unmanned aerial vehicles (UAVs), the detection and identification of UAVs is a vital security issue for the safety of airspace and ground facilities in the no-fly zone. Telemetry radios are important wireless communication devices for UAVs, especially in UAVs beyond the visual line of sight (BVLOS) operating mode. This work focuses on the UAV identification approach using transient signals from UAV telemetry radios instead of the signals from UAV controllers that the former research work depended on. In our novel UAV Radio Frequency (RF) identification system framework based on telemetry radio signals, the EC−α algorithm is optimized to detect the starting point of the UAV transient signal and the detection accuracy at different signal-to-noise ratios (SNR) is evaluated. In the training stage, the Convolutional Neural Network (CNN) model is trained to extract features from raw I/Q data of the transient signals with different waveforms. Its architecture and hyperparameters are analyzed and optimized. In the identification stage, the extracted transient signals are clustered through the Self-Organizing Map (SOM) algorithm and the Clustering Signals Joint Identification (CSJI) algorithm is proposed to improve the accuracy of RF fingerprint identification. To evaluate the performance of our proposed approach, we design a testbed, including two UAVs as the flight platform, a Universal Software Radio Peripheral (USRP) as the receiver, and 20 telemetry radios with the same model as targets for identification. Indoor test results show that the optimized identification approach achieves an average accuracy of 92.3% at 30 dB. In comparison, the identification accuracy of SVM and KNN is 69.7% and 74.5%, respectively, at the same SNR condition. Extensive experiments are conducted outdoors to demonstrate the feasibility of this approach.

Volatilization of benzene on soil surface under different factors: evaluation and modeling

The volatilization of volatile organic compounds following a leakage event is a crucial mechanism that influences their migration and transformation in the soil. It is noteworthy that this process is intricately shaped by soil properties and environmental factors, exhibiting highly complex nonlinear relationships. However, there is currently no reliable mathematical model to predict the nonlinear relationship. To address this gap, the study conducted dynamic experiments considering various factors, including particle size, organic matter content, temperature, wind speed and moisture content. The volatilization rate (k\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$k$$\\end{document}), an important parameter in volatilization kinetics reflecting the speed of volatilization, was calculated by first-order kinetic principle. Finally, an innovative approach was introduced using a Back Propagation Neural Network (BPNN) model for prediction. The findings indicate that wind speed exerts the most significant impact on the volatilization rate of benzene among the examined factors. The application of BPNN demonstrates the model's accuracy in simulating benzene volatilization rates under diverse conditions. The results of K-fold cross-validation alleviate concerns of potential over-prediction, affirming the reliability of the constructed model. This research introduces a novel methodology for predicting volatilization parameters in real-world scenarios.

Performance and Clinical Impact of Radiomics and 3D-CNN Models for the Diagnosis of Neurodegenerative Parkinsonian Syndromes on 18F-FDOPA PET.

The aim of this study was to compare the performance and added clinical value of a semiautomated radiomics model and an automated 3-dimensinal convolutional neural network (3D-CNN) model for diagnosing neurodegenerative parkinsonian syndromes on 18F-FDOPA PET images. This 2-center retrospective study included 687 patients with motor symptoms consistent with parkinsonian syndrome. All patients underwent 18F-FDOPA brain PET scans, acquired on 3 PET systems from 2 different hospitals, and classified as pathological or nonpathological (by an expert nuclear physician). Artificial intelligence models were trained to replicate this medical expert's classification using 2 pipelines. The radiomics pipeline was semiautomated and involved manually segmenting the bilateral caudate and putamen nuclei; 43 radiomic features were extracted and combined using the support vector machine method. The deep learning pipeline was fully automatic and used a 3D-CNN model. Both models were trained on 417 patients and tested on an internal (n = 100) and an external (n = 170) test set. The final models' performance was evaluated using balanced accuracy and compared with that of a junior medical expert and nonexpert nuclear physician. On the internal test set, the 3D-CNN model outperformed the radiomic model with a balanced accuracy of 99% (vs 96%). It led to diagnostic performance similar to that of a junior medical expert (only 1 in 100 patients misclassified by both). On the external test set from a less experienced hospital, the 3D-CNN model allowed physicians to correctly reclassify the diagnosis of 10 out 170 patients (6%). The developed 3D-CNN model can automatically diagnose neurodegenerative parkinsonian syndromes, also reducing diagnostic errors by 6% in less-experienced hospitals.

Gear fault diagnosis based on small channel convolutional neural network under multiscale fusion attention mechanism

AbstractDue to the insufficient feature learning ability and the bloated network structure, the gear fault diagnosis methods based on traditional deep neural networks always suffer from poor diagnosis accuracy and low diagnosis efficiency. Therefore, a small channel convolutional neural network under the multiscale fusion attention mechanism (MSFAM‐SCCNN) is proposed in this paper. First, a small channel convolutional neural network (SCCNN) model is constructed based on the framework of the traditional AlexNet model in order to lightweight the network structure and improve the learning efficiency. Then, a novel multiscale fusion attention mechanism (MSFAM) is embedded into the SCCNN model, which utilizes multiscale striped convolutional windows to extract key features from three dimensions, including temporal, spatial, and channel‐wise, resulting in more precise feature mining. Finally, the performance of the MSFAM‐ SCCNN model is verified using the vibration data of tooth‐broken gears obtained by a self‐designed experimental bench of an ammunition supply and delivery system.

An integrated approach for discrimination of Magnoliae officinalis cortex before and after being processed by ginger juice combining LC/MS, GC/MS, intelligent sensors, and chemometrics.

Magnoliae officinalis cortex (MOC) is an important traditional Chinese medicine (TCM), and both raw and stir-fried MOC were commonly used in clinic. This study aimed to discriminate MOC and MOC stir-fried with ginger juice (MOCG) using an integrated approach combining liquid chromatography/mass spectrometry (LC/MS), gas chromatography/mass spectrometry (GC/MS), intelligent sensors, and chemometrics. The sensory characters of the samples were digitalized using intelligent sensors, i.e., colorimeter, electronic nose, and electronic tongue. Meanwhile, the chemical profiles of the samples were analyzed using LC/MS and GC/MS methods. Chemometric models were constructed to discriminate samples of MOC and MOCG based on not only the sensory data but also the chemical data. The differential sensory characters (L* and b* from colorimeter, ANS from electronic tongue, W1S and W2S from electronic nose) and the differential chemical compounds (26 and 11 compounds from LC/MS and GC/MS, respectively) were discovered between MOC and MOCG. Furthermore, twelve differential compounds showed good relations with differential sensory characters. Finally, artificial neural network models were established to discriminate samples of MOC and MOCG, in which W1S, W2S, ANS, b*, and 10 differential compounds were among the top 10 important variables, respectively. Samples of MOC and MOCG can be discriminated not only by the digitalized data of color, taste, and scent detected by intelligent sensors but also by chemical information obtained from LC/MS and GC/MS using chemometrics. The variations in sensory characters and chemical compounds between MOC and MOCG partially resulted from the Maillard reaction products and the oxidation of some compounds in the stir-frying process.

Explainable Artificial Intelligence Approach for Improving Head-Mounted Fault Display Systems

To fully harness the potential of wind turbine systems and meet high power demands while maintaining top-notch power quality, wind farm managers run their systems 24 h a day/7 days a week. However, due to the system’s large size and the complex interactions of its many components operating at high power, frequent critical failures occur. As a result, it has become increasingly important to implement predictive maintenance to ensure the continued performance of these systems. This paper introduces an innovative approach to developing a head-mounted fault display system that integrates predictive capabilities, including deep learning long short-term memory neural networks model integration, with anomaly explanations for efficient predictive maintenance tasks. Then, a 3D virtual model, created from sampled and recorded data coupled with the deep neural diagnoser model, is designed. To generate a transparent and understandable explanation of the anomaly, we propose a novel methodology to identify a possible subset of characteristic variables for accurately describing the behavior of a group of components. Depending on the presence and risk level of an anomaly, the parameter concerned is displayed in a piece of specific information. The system then provides human operators with quick, accurate insights into anomalies and their potential causes, enabling them to take appropriate action. By applying this methodology to a wind farm dataset provided by Energias De Portugal, we aim to support maintenance managers in making informed decisions about inspection, replacement, and repair tasks.

Domain adaptation of transformer-based neural network model for clinical note classification in Indian healthcare

Domain adaptation of transformer-based neural network model for clinical note classification in Indian healthcare

A Novel Hybrid Model for Financial Forecasting Based on CEEMDAN-SE and ARIMA-CNN-LSTM

Financial time series data are characterized by non-linearity, non-stationarity, and stochastic complexity, so predicting such data presents a significant challenge. This paper proposes a novel hybrid model for financial forecasting based on CEEMDAN-SE and ARIMA- CNN-LSTM. With the help of the CEEMDAN-SE method, the original data are decomposed into several IMFs and reconstructed via sample entropy into a lower-complexity stationary high-frequency component and a low-frequency component. The high-frequency component is predicted by the ARIMA statistical forecasting model, while the low-frequency component is predicted by a neural network model combining CNN and LSTM. Compared to some classical prediction models, our algorithm exhibits superior performance in terms of three evaluation indexes, namely, RMSE, MAE, and MAPE, effectively enhancing model accuracy while reducing computational overhead.

YOLO-HyperVision: A vision transformer backbone-based enhancement of YOLOv5 for detection of dynamic traffic information

With the increase of traffic flow in modern urban areas, traffic congestion has become a serious problem that affects people’s normal production and life. Using target detection technology instead of manual labor can quickly detect the road traffic situation and provide timely information about the traffic flow. However, when using drones to observe the traffic flow in the air, the perspective effect will cause the detected vehicles and pedestrians to be very small, and the scale difference between different categories of targets is large, which increases the detection difficulty of a single convolutional neural network model. In order to solve the problem of low accuracy of traditional single-stage target detection models, this study proposes an improved Yolov5 vehicle target detection model with Vision Transformer (VIT) backbone, You Only Look Once-HyperVision (YOLO-HV), which aims to solve the problem of poor multi-scale target recognition performance caused by the inability of traditional CNN networks to integrate contextual information, and help drones achieve more efficient and accurate traffic flow recognition functions. This study deeply integrates the Vision Transformer (VIT) backbone and Convolutional Neural Network (CNN), effectively combining the multi-scale detection advantages of Vision Transformer and the inductive bias ability of Convolutional Neural Network, and adds multi-scale residual modules and context correlation enhancement modules, which greatly improves the recognition accuracy of single-stage detectors for drone images. Through comparative experiments on the VisDrone dataset, it is found that the detection performance of this model is improved compared with several commonly used detection models. YOLO-HV can increase the mean average precision (mAP) by 3.3% compared with the pure convolutional network of the same model size. YOLO-HV model has achieved excellent performance in the task of traffic flow image detection taken by drones, and can more accurately identify and classify road vehicles than various target detection models.