V2 Model Research Articles

DiaNet v2 deep learning based method for diabetes diagnosis using retinal images

Diabetes mellitus (DM) is a prevalent chronic metabolic disorder linked to increased morbidity and mortality. With a significant portion of cases remaining undiagnosed, particularly in the Middle East North Africa (MENA) region, more accurate and accessible diagnostic methods are essential. Current diagnostic tests like fasting plasma glucose (FPG), oral glucose tolerance tests (OGTT), random plasma glucose (RPG), and hemoglobin A1c (HbA1c) have limitations, leading to misclassifications and discomfort for patients. The aim of this study is to enhance diabetes diagnosis accuracy by developing an improved predictive model using retinal images from the Qatari population, addressing the limitations of current diagnostic methods. This study explores an alternative approach involving retinal images, building upon the DiaNet model, the first deep learning model for diabetes detection based solely on retinal images. The newly proposed DiaNet v2 model is developed using a large dataset from Qatar Biobank (QBB) and Hamad Medical Corporation (HMC) covering wide range of pathologies in the the retinal images. Utilizing the most extensive collection of retinal images from the 5545 participants (2540 diabetic patients and 3005 control), DiaNet v2 is developed for diabetes diagnosis. DiaNet v2 achieves an impressive accuracy of over 92%, 93% sensitivity, and 91% specificity in distinguishing diabetic patients from the control group. Given the high prevalence of diabetes and the limitations of existing diagnostic methods in clinical setup, this study proposes an innovative solution. By leveraging a comprehensive retinal image dataset and applying advanced deep learning techniques, DiaNet v2 demonstrates a remarkable accuracy in diabetes diagnosis. This approach has the potential to revolutionize diabetes detection, providing a more accessible, non-invasive and accurate method for early intervention and treatment planning, particularly in regions with high diabetes rates like MENA.

Big data clustering using fuzzy based energy efficient clustering and MobileNet V2

Big data analytics (BDA) is a systematic way to analyze and detect various patterns, relationships, and trends in vast amounts of data. Big data analysis and processing require significant effort, techniques, and equipment. The Hadoop framework software uses the MapReduce approach to do large-scale data analysis using parallel processing in order to generate results as soon as possible. Due to the traditional algorithm’s longer execution time and difficulty in processing big amounts of data, this is one of the main issues. Clusters are highly correlated inside each other but are not highly correlated with one another. The technique of effectively allocating limited resources is known as an optimization algorithm for clustering. For processing large amounts of data with several dimensions, the conventional optimization approach is insufficient. By using a fuzzy method, this can be prevented. In this paper, we proposed Fuzzy based energy efficient clustering approach to enhance the clustering mechanism. In summary, Fuzzy based energy efficient clustering introduces a function that measures the distance between the cluster center and the instance, which aids in improved clustering, and we then present the MobileNet V2 model to improve efficiency and speed up computation. To enhance the method’s performance and reduce its time complexity, the distributed database simulates the shared memory space and parallelizes on the MapReduce framework on the Hadoop cloud computing platform. The proposed approach is evaluated using performance metrics such as Accuracy, Precision, Adjusted Rand Index (ARI), Recall, F1-Score, and Normalized Mutual Information (NMI). The experimental findings indicate that the proposed approach outperforms the existing techniques in terms of clustering accuracy.

An effective object detection and tracking using automated image annotation with inception based faster R-CNN model

The present study advances object detection and tracking techniques by proposing a novel model combining Automated Image Annotation with Inception v2-based Faster RCNN (AIA-IFRCNN). The research methodology utilizes the DCF-CSRT model for image annotation, Faster RCNN for object detection, and the inception v2 model for feature extraction, followed by a softmax layer for image classification. The proposed AIA-IFRCNN model is evaluated on three benchmark datasets: Bird (Dataset 1), UCSDped2 (Dataset 2), and Under Water (Dataset 3), to determine prediction accuracy, annotation time, Center Location Error (CLE), and Overlap Rate (OR). The experimental results indicate that the AIA-IFRCNN model outperformed existing models regarding detection accuracy and tracking performance. Notably, it achieved a maximum detection accuracy of 95.62 % on Dataset 1, outperforming other models. Additionally, it achieved minimum average CLE values of 4.16, 5.78, and 3.54, and higher overlap rates of 0.92, 0.90, and 0.94 on the respective datasets (1, 2 and 3). Hence, this research work on object detection and tracking using the AIA-IFRCNN model is essential for improving system efficiency and fostering innovation in the field of computer vision and object tracking.

Automated Lung and Colon Cancer Classification Using Histopathological Images.

Cancer is the leading cause of mortality in the world. And among all cancers lung and colon cancers are 2 of the most common causes of death and morbidity. The aim of this study was to develop an automated lung and colon cancer classification system using histopathological images. An automated lung and colon classification system was developed using histopathological images from the LC25000 dataset. The algorithm development included data splitting, deep neural network model selection, on the fly image augmentation, training and validation. The core of the algorithm was a Swin Transform V2 model, and 5-fold cross validation was used to evaluate model performance. The model performance was evaluated using Accuracy, Kappa, confusion matrix, precision, recall, and F1. Extensive experiments were conducted to compare the performances of different neural networks including both mainstream convolutional neural networks and vision transformers. The Swin Transform V2 model achieved a 1 (100%) on all metrics, which is the first single model to obtain perfect results on this dataset. The Swin Transformer V2 model has the potential to be used to assist pathologists in classifying lung and colon cancers using histopathology images.

Enhancing Diffusion Models with 3D Perspective Geometry Constraints

While perspective is a well-studied topic in art, it is generally taken for granted in images. However, for the recent wave of high-quality image synthesis methods such as latent diffusion models, perspective accuracy is not an explicit requirement. Since these methods are capable of outputting a wide gamut of possible images, it is difficult for these synthesized images to adhere to the principles of linear perspective. We introduce a novel geometric constraint in the training process of generative models to enforce perspective accuracy. We show that outputs of models trained with this constraint both appear more realistic and improve performance of downstream models trained on generated images. Subjective human trials show that images generated with latent diffusion models trained with our constraint are preferred over images from the Stable Diffusion V2 model 70% of the time. SOTA monocular depth estimation models such as DPT and PixelFormer, fine-tuned on our images, outperform the original models trained on real images by up to 7.03% in RMSE and 19.3% in SqRel on the KITTI test set for zero-shot transfer.

Convert-Pheno: A software toolkit for the interconversion of standard data models for phenotypic data

Efficient sharing and integration of phenotypic data is crucial for advancing biomedical research and enhancing patient outcomes in precision medicine and public health. To achieve this, the health data community has developed standards to promote the harmonization of variable names and values. However, the use of diverse standards across different research centers can hinder progress. Here we present Convert-Pheno, an open-source software toolkit that enables the interconversion of common data models for phenotypic data such as Beacon v2 Models, CDISC-ODM, OMOP-CDM, Phenopackets v2, and REDCap. Along with the software, we have created a detailed documentation that includes information on deployment and installation.

Extending the Breast Cancer Surveillance Consortium Model of Invasive Breast Cancer.

We extended the Breast Cancer Surveillance Consortium (BCSC) version 2 (v2) model of invasive breast cancer risk to include BMI, extended family history of breast cancer, and age at first live birth (version 3 [v3]) to better inform appropriate breast cancer prevention therapies and risk-based screening. We used Cox proportional hazards regression to estimate the age- and race- and ethnicity-specific relative hazards for family history of breast cancer, breast density, history of benign breast biopsy, BMI, and age at first live birth for invasive breast cancer in the BCSC cohort. We evaluated calibration using the ratio of expected-to-observed (E/O) invasive breast cancers in the cohort and discrimination using the area under the receiver operating characteristic curve (AUROC). We analyzed data from 1,455,493 women age 35-79 years without a history of breast cancer. During a mean follow-up of 7.3 years, 30,266 women were diagnosed with invasive breast cancer. The BCSC v3 model had an E/O of 1.03 (95% CI, 1.01 to 1.04) and an AUROC of 0.646 for 5-year risk. Compared with the v2 model, discrimination of the v3 model improved most in Asian, White, and Black women. Among women with a BMI of 30.0-34.9 kg/m2, the true-positive rate in women with an estimated 5-year risk of 3% or higher increased from 10.0% (v2) to 19.8% (v3) and the improvement was greater among women with a BMI of ≥35 kg/m2 (7.6%-19.8%). The BCSC v3 model updates an already well-calibrated and validated breast cancer risk assessment tool to include additional important risk factors. The inclusion of BMI was associated with the largest improvement in estimated risk for individual women.

The suitability of using Uniswap V2 model to analyze V3 data

Decentralized exchanges’ popularity is rising, with liquidity pools widely used for trading. Uniswap V3, a newer version, offers advanced features, but it is more complex to analyze compared to V1 and V2. We compared a simple V2 model’s theoretical predictions with Uniswap V3 data. Surprisingly, the V2 model accurately predicted the V3 data in 97.1% of transactions, with a deviation of less than 0.1%. Accuracy was higher in active pools with substantial transaction volume and liquidity, while inactive pools performed less effectively. This approach aids researchers in assessing V2 model suitability for Uniswap V3 data analysis.

An AI Dietitian for Type 2 Diabetes Mellitus Management Based on Large Language and Image Recognition Models: Preclinical Concept Validation Study.

Nutritional management for patients with diabetes in China is a significant challenge due to the low supply of registered clinical dietitians. To address this, an artificial intelligence (AI)-based nutritionist program that uses advanced language and image recognition models was created. This program can identify ingredients from images of a patient's meal and offer nutritional guidance and dietary recommendations. The primary objective of this study is to evaluate the competence of the models that support this program. The potential of an AI nutritionist program for patients with type 2 diabetes mellitus (T2DM) was evaluated through a multistep process. First, a survey was conducted among patients with T2DM and endocrinologists to identify knowledge gaps in dietary practices. ChatGPT and GPT 4.0 were then tested through the Chinese Registered Dietitian Examination to assess their proficiency in providing evidence-based dietary advice. ChatGPT's responses to common questions about medical nutrition therapy were compared with expert responses by professional dietitians to evaluate its proficiency. The model's food recommendations were scrutinized for consistency with expert advice. A deep learning-based image recognition model was developed for food identification at the ingredient level, and its performance was compared with existing models. Finally, a user-friendly app was developed, integrating the capabilities of language and image recognition models to potentially improve care for patients with T2DM. Most patients (182/206, 88.4%) demanded more immediate and comprehensive nutritional management and education. Both ChatGPT and GPT 4.0 passed the Chinese Registered Dietitian examination. ChatGPT's food recommendations were mainly in line with best practices, except for certain foods like root vegetables and dry beans. Professional dietitians' reviews of ChatGPT's responses to common questions were largely positive, with 162 out of 168 providing favorable reviews. The multilabel image recognition model evaluation showed that the Dino V2 model achieved an average F1 score of 0.825, indicating high accuracy in recognizing ingredients. The model evaluations were promising. The AI-based nutritionist program is now ready for a supervised pilot study.

Recognition and statistical method of cows rumination and eating behaviors based on Tensorflow.js

Information about dairy cow ruminating is closely associated with the health status of dairy cows. Therefore, it is of great significance to recognize and make statistics of dairy cows’ ruminating and feeding behavior. Concerning conventional recognition methods which are dependent on contact type devices, they have some defects of poor instantaneity and strong stress responses. As for recognition based on machine vision, it needs to transmit masses of data and raises high requirements for the cloud server and network performance. According to principles of edge computing, the model is deployed via Tensorflow.js in an edge device in the present study, constructing a recognition and statistical system for ruminating and feeding behavior of dairy cows. Through the application programming interface (API) of the browser, an edge device is able to invoke a camera and acquire dairy cow images. Then, the images can be inputted in the SSD MobileNet V2 model, which is followed by inference based on browser hashrate. Moreover, the edge device merely uploads recognition results to the cloud server for statistics, which features high instantaneity and compatibility. In terms of recognizing ruminating and feeding behavior of dairy cows, the proposed system has a precision ratio of 96.50%, a recall rate of 91.77%, an F1-score of 94.08%, specificity of 91.36%, and accuracy of 91.66%. This suggests that the proposed method is effective in recognizing dairy cow behavior.

Improved deep learning image classification algorithm based on Swin Transformer V2

While convolutional operation effectively extracts local features, their limited receptive fields make it challenging to capture global dependencies. Transformer, on the other hand, excels at global modeling and effectively captures global dependencies. However, the self-attention mechanism used in Transformers lacks a local mechanism for information exchange within specific regions. This article attempts to leverage the strengths of both Transformers and convolutional neural networks (CNNs) to enhance the Swin Transformer V2 model. By incorporating both convolutional operation and self-attention mechanism, the enhanced model combines the local information-capturing capability of CNNs and the long-range dependency-capturing ability of Transformers. The improved model enhances the extraction of local information through the introduction of the Swin Transformer Stem, inverted residual feed-forward network, and Dual-Branch Downsampling structure. Subsequently, it models global dependencies using the improved self-attention mechanism. Additionally, downsampling is applied to the attention mechanism’s Q and K to reduce computational and memory overhead. Under identical training conditions, the proposed method significantly improves classification accuracy on multiple image classification datasets, showcasing more robust generalization capabilities.

Localization model and rank-based features selection approach for the classification of GGO and consolidation stages of COVID-19

Diseases like Middle East respiratory syndrome (MERS) and severe acute respiratory syndrome (SARS) have been caused by a large family virus known as Coronavirus. High-resolution computed tomography (HRCT) chest is used as a useful tool for the triage and stratification of COVID-19-suspected patients having respiratory symptoms until they are confirmed by polymerase chain reaction (PCR) test. To further expedite the reporting of HRCT chest in a short time, computer-aided detection algorithms can be more useful. In this work, a method is proposed that consists of three phases including COVID-19 classification, localization, and segmentation. The COVID-19-ResNet-50 model is designed for COVID-19 classification using HRCT slices. The features are obtained from a pre-trained ResNet-50 model, and the process of identifying the most pertinent features is carried out using the Wilcoxon Separability Criteria (WSC). The linear discriminant (LD) classifier is used to classify COVID-19 providing an accuracy of 0.99. To localize the COVID-19-infected region, the classified HRCT slices are fed as input to an open neural network exchange (ONNX) that is a backbone of the Tiny-You Look at Once (YOLO)-v2 model. The proposed COVID-19-YOLOv2 model localized the COVID-19 lesions with 0.948 intersections over union (IoU). A three-dimensional semantic neural network comprising 10 selected layers is proposed whose training based on the best-selected hyperparameters is performed, it accurately segments the localized infected regions having an 0.98 IoU. The proposed method is accessed on national and international datasets that provide improved performance as compared to the most recent works published in this domain.

Artificial Intelligence-Driven Eye Disease Classification Model

Eye diseases can result in various challenges and visual impairments. These diseases can affect an individual’s quality of life and general health and well-being. The symptoms of eye diseases vary widely depending on the nature and severity of the disease. Early diagnosis can protect individuals from visual impairment. Artificial intelligence (AI)-based eye disease classification (EDC) assists physicians in providing effective patient services. However, the complexities of the fundus image affect the classifier’s performance. There is a demand for a practical EDC for identifying eye diseases in the earlier stages. Thus, the author intends to build an EDC model using the deep learning (DL) technique. Denoising autoencoders are used to remove the noises and artifacts from the fundus images. The single-shot detection (SSD) approach generates the key features. The whale optimization algorithm (WOA) with Levy Flight and Wavelet search strategy is followed for selecting the features. In addition, the Adam optimizer (AO) is applied to fine-tune the ShuffleNet V2 model to classify the fundus images. Two benchmark datasets, ocular disease intelligent recognition (ODIR) and EDC datasets, are utilized for performance evaluation. The proposed EDC model achieved accuracy and Kappa values of 99.1 and 96.4, and 99.4 and 96.5, in the ODIR and EDC datasets, respectively. It outperformed the recent EDC models. The findings highlight the significance of the proposed EDC model in classifying eye diseases using complex fundus images. Healthcare centers can implement the proposed model to improve their standards and serve a more significant number of patients. In the future, the proposed model can be extended to identify a comprehensive range of eye diseases.

Optimizing Generative Adversarial Network (GAN) Models for Non-Pneumatic Tire Design

Pneumatic tires are used in diverse industries. However, their design is difficult, as it relies on the knowledge of experienced designers. In this paper, we generate images of non-pneumatic tire designs with patterns based on shapes and lines for different generative adversarial network (GAN) models and test the performance of the models. Using OpenCV, 2000 training images were generated, corresponding to spoke, curve, triangle, and honeycomb non-pneumatic tires. The images created for training were used after removing highly similar images by applying mean squared error (MSE) and structural similarity index (SSIM). To identify the best model for generating patterns of regularly shaped non-pneumatic tires, GAN, deep convolutional generative adversarial network (DCGAN), StarGAN v2, StyleGAN v2-ADA, and ProjectedGAN were compared and analyzed. In the qualitative evaluation, the GAN, DCGAN, StarGAN v2, and StyleGAN v2-ADA models distorted the circle shape and did not maintain a consistent pattern, but ProjectedGAN retained consistency in the circle, and the pattern was less distorted than in the other GAN models. When evaluating quantitative metrics, ProjectedGAN performed the best among several techniques when the difference between the generated and actual image distributions was measured.

Identification of Individual Hanwoo Cattle by Muzzle Pattern Images through Deep Learning.

The objective of this study was to identify Hanwoo cattle via a deep-learning model using muzzle images. A total of 9230 images from 336 Hanwoo were used. Images of the same individuals were taken at four different times to avoid overfitted models. Muzzle images were cropped by the YOLO v8-based model trained with 150 images with manual annotation. Data blocks were composed of image and national livestock traceability numbers and were randomly selected and stored as train, validation test data. Transfer learning was performed with the tiny, small and medium versions of Efficientnet v2 models with SGD, RMSProp, Adam and Lion optimizers. The small version using Lion showed the best validation accuracy of 0.981 in 36 epochs within 12 transfer-learned models. The top five models achieved the best validation accuracy and were evaluated with the training data for practical usage. The small version using Adam showed the best test accuracy of 0.970, but the small version using RMSProp showed the lowest repeated error. Results with high accuracy prediction in this study demonstrated the potential of muzzle patterns as an identification key for individual cattle.

Melanoma Espial Employing Deep Learning Applied to Mobilenet

Abstract: The most recent advancements in smartphone-based skin cancer diagnosis software enable convenient, portable methods for melanoma risk assessment and diagnosis for skin cancer early detection due to the issue of trade-offs. [12] The majority of skin cancer detection applications do the image analysis on the server due to the (time complexity and error rate) limitations of running a machine learning algorithm for image analysis on a smartphone. In this work, we use the MobileNet v2 deep learning model to examine the performance of skin cancer picture identification and classification on Android devices. [10] We evaluate the effectiveness of a number of factors, including parameter setup, picture acquisition technique, computer and Android-based image analysis, and object identification and classification methods. Melanoma and skin cancer are utilized to evaluate the effectiveness of the suggested approach. [15] The measurement uses the testing techniques' running times, sensitivity, accuracy, and specificity. The optimum parameter for the MobileNet v2 model on Android utilizing photos from the smartphone camera gives 91.346% accuracy for classification, according to the experiment's findings. The Android app's ability to recognize objects and classify them made it possible to analyze skin cancer. Android-based image analysis maintains the computation time threshold that indicates user comfort and matches the computer's performance accuracy for high-quality photos. [18] These results drove the creation of an Android app for disease detection processing that makes use of the camera on smartphones and seeks to achieve high-accuracy real-time detection and classification.

Verification of Some Weather Elements Forecasts from BSC-DREAM8b V2 Model

Abstract The aim of this study is to verify 2-meter temperature and mean sea level pressure forecasts from BSC-DREAM8B V2.0 model, which is a short-term numerical forecasting model against observed data from the Iraqi Meteorological Organization & Seismology constitution and modelled data (ERA-Interim) reanalysis from the European Centre for Medium Range Weather Forecasts (ECMWF) for the period 1-3 June 2012 for four selected stations in Iraq, they are Mosul, Baghdad, Rutba and Basra which represent northern, central, western and southern of Iraq respectively by using some statistical methods including Mean Absolute Error (MAE), Bias Error (BE) and Correlation Coefficient (R). The results showed that the BSC-DREAM8B V2.0 model worked well in representing the daily temperature variation and with relatively good compatibility with the ERA-I modelled data, however, the latter showed lower absolute error rates. The two modelled data showed the highest absolute error rate for Baghdad station whereas BSC showed lower values for Basra station and different values for Rutba and Mosul, and the predicted values were overestimated for all stations and for both models except for Rutba station where ERA-I data showed underestimated value. As for the surface pressure at mean sea level, the BSC-DREAM8B V2.0 model showed its ability to simulate the daily change of pressure and worked in a good consistency with ERA-I modelled data with few hetopascal as mean absolute error. Both modelled data showed underestimated values for all stations, and thus, BSC modelled data is better for simulating temperature and mean sea level pressure in low elevated regions compared to ERA-Interim data and vice versa.

Development of a Forest Fire Detection System Using a Drone-based Convolutional Neural Network Model

Considering forest fires cause environmental destruction, ecosystem collapse, and severe damage to human lives and nature, developing a real-time, accurate, and stable forest fire detection system has become a critical issue in modern society. In this study, a drone-based forest fire detection system was developed using a convolutional neural network (CNN) model. Real-time forest fire detection models were developed using the CNN-based MobileNet algorithm, and their fire detection performance was evaluated. The main research results indicated that errors decreased and accuracy tended to increase during the model training and validation process as training progressed. Moreover, the V1 model exhibited the highest validation accuracy of 0.9466 among the MobileNet V1, V2, and V3 models and showed the highest accuracy of 0.9667 in evaluating the new test dataset during the model evaluation process.

Lane Reconstruction for Self-Driving Vehicles on Dynamic Road Networks

Lane detection is a crucial task that involves identifying lane lines and proper markings on road structures. The field of lane detection has gained significant attention due to the increasing use of autonomous vehicles and ADAS systems, as they require robust lane detection systems to navigate safely in complex environments. While previous studies have focused on lane detection in various scenarios, lane line reconstruction for dynamic road networks has been largely overlooked. To address this gap, we propose a novel approach to reconstruct lane lines for dynamic road networks using roadblocks, specifically red and white striped markers, as cues. To evaluate the effectiveness of our proposed approach, we first built and labeled our own dataset by extracting 1854 frames from recorded videos. We trained a Faster-RCNN V2 model on 80% of the dataset to detect the roadblocks and used the OpenCV DNN module, along with a range of techniques, such as ROI selection, color selection, region masking, and polynomial and linear fittings for lane detection and reconstruction. Our approach successfully reconstructed the left, middle, and right lane lines, achieving an overall accuracy of 68%, a recall of 73%, and a precision score of 73% on an unseen set of 50 images. Our study contributes to the field by proposing a novel approach for reconstructing lane lines in dynamic road networks using roadblocks as markers. This approach has the potential to enhance the accuracy and robustness of lane detection systems in complex environments. Furthermore, our work addresses a gap in the existing literature and provides insights for future research in this area.

The First Study of White Rust Disease Recognition by Using Deep Neural Networks and Raspberry Pi Module Application in Chrysanthemum

Growth factors affect farm owners, environmental conditions, nutrient adaptation, and resistance to chrysanthemum diseases. Healthy chrysanthemum plants can overcome all these factors and provide farms owners with a lot of income. Chrysanthemum white rust disease is a common disease that occurs worldwide; if not treated promptly, the disease spreads to the entire leaf surface, causing the plant’s leaves to burn, turn yellow, and fall prematurely, reducing the photosynthetic performance of the plant and the appearance of the flower branches. In Korea, chrysanthemum white rust disease most often occurs during the spring and autumn seasons, when temperature varies during the summer monsoon, and when ventilation is poor in the winter. Deep neural networks were used to determine healthy and unhealthy plants. We applied the Raspberry Pi 3 module to recognize white rust and test four neural network models. The five main deep neural network processes utilized for a dataset of non-diseased and white rust leaves include: (1) data collection; (2) data partitioning; (3) feature extraction; (4) feature engineering; and (5) prediction modeling based on the train–test loss of 35 epochs within 20 min using Linux. White rust recognition is performed for comparison using four models, namely, DenseNet121, ResNet50, VGG-19, and MobileNet v2. The qualitative white rust detection system is achieved using a Raspberry Pi 3 module. All models accomplished an accuracy of over 94%, and MobileNet v2 achieved the highest accuracy, precision, and recall at over 98%. In the precision comparison, DenseNet121 obtained the second highest recognition accuracy of 97%, whereas ResNet50 and VGG-19 achieved slightly lower accuracies at 95% and 94%, respectively. Qualitative results were obtained using the Raspberry Pi 3 module to assess the performance of the four models. All models had accuracies of over 91%, with ResNet50 obtaining a value of 91%, VGG19 reaching a value of 93%, and DenseNet121 reaching 95%. The highest accuracy rate was 97% (MobileNet v2). MobileNet v2 was validated as the most effective model to recognize white rust in chrysanthemums using the Raspberry Pi 3 system. Raspberry Pi 3 module was considered, in conjunction with the MobileNet v2 model, to be the best application system. MobileNet v2 and Raspberry Pi require a low cost for the recognition of chrysanthemum white rust and the diagnosis of chrysanthemum plant health conditions, reducing the risk of white rust disease and minimizing costs and efforts while improving floral production. Chrysanthemum farmers should consider applying the Raspberry Pi module for detecting white rust, protecting healthy plant growth, and increasing yields with low-cost.