Deep Learning Classification Model Research Articles

Hydrogen storage metal-organic framework classification models based on crystal graph convolutional neural networks

Metal-organic frameworks (MOFs) have been considered as promising physical adsorbents for hydrogen storage due to their high porosity and structural tunability. We selected 7643 real MOFs from the computation-ready MOF 2019 database to screen high-performance materials for hydrogen storage based on the grand canonical Monte Carlo (GCMC) simulations. Based on the obtained data set, we proposed a deep learning classification model powered by the crystal graph convolutional neural networks (CGCNN) for the discovery of the optimal hydrogen storage MOFs. It is demonstrated that our classification model based on CGCNN algorithms exhibits a high prediction accuracy with the area under the ROC-plot curve (AUC) of 0.9208 for hydrogen storage performance of volumetric deliverable capacity (VDC). Compared with the classification models based on other machine learning algorithms such as random forest, our CGCNN model demonstrates the advantages of fast prediction with no feature extraction and little accuracy loss. When the trained CGCNN model was used to predict the classification of the unfamiliar samples (the randomly selected 1000 MOFs from the 137,953 hypothetical MOF database), we also obtain high accuracy with AUC > 0.81, indicating that this model exhibits reliable transferability for other types of MOFs. Meanwhile, we also elucidated the relationship between structure and performance of MOFs for hydrogen storage using the decision tree algorithms and quantitative structure–property analysis. Furthermore, the hydrogen adsorption performance and mechanism of top-performance MOFs were analyzed by adsorption isotherms, radial distribution functions, and mass center density distribution of equilibrium configurations. All those insights from atomic simulations and machine learnings can accelerate the discovery of new nanoporous materials not only for gas adsorption in MOFs but also for gas separation in other types of porous materials.

FCF: Feature complement fusion network for detecting COVID-19 through CT scan images

COVID-19 spreads and contracts people rapidly, to diagnose this disease accurately and timely is essential for quarantine and medical treatment. RT-PCR plays a crucial role in diagnosing the COVID-19, whereas computed tomography (CT) delivers a faster result when combining artificial assistance. Developing a Deep Learning classification model for detecting the COVID-19 through CT images is conducive to assisting doctors in consultation. We proposed a feature complement fusion network (FCF) for detecting COVID-19 through lung CT scan images. This framework can extract both local features and global features by CNN extractor and ViT extractor severally, which successfully complement the deficiency problem of the receptive field of the other. Due to the attention mechanism in our designed feature complement Transformer (FCT), extracted local and global feature embeddings achieve a better representation. We combined a supervised with a weakly supervised strategy to train our model, which can promote CNN to guide the VIT to converge faster. Finally, we got a 99.34% accuracy on our test set, which surpasses the current state-of-art popular classification model. Moreover, this proposed structure can easily extend to other classification tasks when changing other proper extractors.

Class Incremental Learning With Few-Shots Based on Linear Programming for Hyperspectral Image Classification.

Hyperspectral imaging (HSI) classification has drawn tremendous attention in the field of Earth observation. In the big data era, explosive growth has occurred in the amount of data obtained by advanced remote sensors. Inevitably, new data classes and refined categories appear continuously, and such data are limited in terms of the timeliness of application. These characteristics motivate us to build an HSI classification model that learns new classifying capability rapidly within a few shots while maintaining good performance on the original classes. To achieve this goal, we propose a linear programming incremental learning classifier (LPILC) that can enable existing deep learning classification models to adapt to new datasets. Specifically, the LPILC learns the new ability by taking advantage of the well-trained classification model within one shot of the new class without any original class data. The entire process requires minimal new class data, computational resources, and time, thereby making LPILC a suitable tool for some time-sensitive applications. Moreover, we utilize the proposed LPILC to implement fine-grained classification via the well-trained original coarse-grained classification model. We demonstrate the success of LPILC with extensive experiments based on three widely used hyperspectral datasets, namely, PaviaU, Indian Pines, and Salinas. The experimental results reveal that the proposed LPILC outperforms state-of-the-art methods under the same data access and computational resource. The LPILC can be integrated into any sophisticated classification model, thereby bringing new insights into incremental learning applied in HSI classification.

IsVoNet8: A Proposed Deep Learning Model for Classification of Some Fish Species

Fish is a very diverse animal of all vertebrate animal groups, of which there are more than 33,000 species in the world. There are different types of fish in the four major geographical regions of Turkey. Classification of different fish species is very important for aquaculture, stock management of water bodies, monitoring of aquatic organisms and conservation of marine biology. In the classification of fish, both knowledge and great effort are required to determine the characteristics of fish. Traditionally, however, manual classification of extrinsic characteristics of different fish species has been a difficult and time-consuming process due to their close resemblance to each other. Recently, deep learning methods used in the light of developments in the field of computer vision have facilitated the training of fish image classification models and the recognition of various fish species. In this study, a new evolutionary neural network model classifying 8 different fish species using deep learning methods was proposed. The proposed model is compared with the ResNet-50, ResNet-101 and VGG16 models. The success accuracies obtained as a result of the comparison are respectively; 98.12% in the proposed model, 91.37% in the ResNet-50 model, 86.12% in the ResNet-101 model and 97.75% in the VGG16 model. It has been observed that the proposed model classifies sea fish, which is widely consumed in our country, with higher performance compared to other models.

Integrating content-based image retrieval and deep learning to improve wafer bin map defect patterns classification

ABSTRACT Defect dies scattering on semiconductor wafer bin maps (WBM) tends to form specific patterns that point to particular manufacturing problems. The distribution of defect patterns from the shop floor is often highly imbalanced, leading to the challenge of having insufficient data about defect pattern types when building deep learning classification models. The method for completing such analysis in a timely manner with limited data is of critical interest. This study developed a method for applying content-based image retrieval (CBIR) and convolutional neural networking (CNN) to WBM defect patterns classification to solve the data imbalance problem and to improve accuracy when using relatively a small quantity of data. In this research, 3,600 WBMs featuring 12 defect pattern types were selected from the WM-811 K dataset for empirical validation. Using only 1,400 CNN training data elements, the overall classification accuracy reached 98.44%.

Deep learning model for the automatic classification of COVID-19 pneumonia, non-COVID-19 pneumonia, and the healthy: a multi-center retrospective study

This retrospective study aimed to develop and validate a deep learning model for the classification of coronavirus disease-2019 (COVID-19) pneumonia, non-COVID-19 pneumonia, and the healthy using chest X-ray (CXR) images. One private and two public datasets of CXR images were included. The private dataset included CXR from six hospitals. A total of 14,258 and 11,253 CXR images were included in the 2 public datasets and 455 in the private dataset. A deep learning model based on EfficientNet with noisy student was constructed using the three datasets. The test set of 150 CXR images in the private dataset were evaluated by the deep learning model and six radiologists. Three-category classification accuracy and class-wise area under the curve (AUC) for each of the COVID-19 pneumonia, non-COVID-19 pneumonia, and healthy were calculated. Consensus of the six radiologists was used for calculating class-wise AUC. The three-category classification accuracy of our model was 0.8667, and those of the six radiologists ranged from 0.5667 to 0.7733. For our model and the consensus of the six radiologists, the class-wise AUC of the healthy, non-COVID-19 pneumonia, and COVID-19 pneumonia were 0.9912, 0.9492, and 0.9752 and 0.9656, 0.8654, and 0.8740, respectively. Difference of the class-wise AUC between our model and the consensus of the six radiologists was statistically significant for COVID-19 pneumonia (p value = 0.001334). Thus, an accurate model of deep learning for the three-category classification could be constructed; the diagnostic performance of our model was significantly better than that of the consensus interpretation by the six radiologists for COVID-19 pneumonia.

A Deep Learning Model for Classification of Parotid Neoplasms Based on Multimodal Magnetic Resonance Image Sequences.

To design a deep learning model based on multimodal magnetic resonance image (MRI) sequences for automatic parotid neoplasm classification, and to improve the diagnostic decision-making in clinical settings. First, multimodal MRI sequences were collected from 266 patients with parotid neoplasms, and an artificial intelligence (AI)-based deep learning model was designed from scratch, combining the image classification network of Resnet and the Transformer network of Natural language processing. Second, the effectiveness of the deep learning model was improved through the multi-modality fusion of MRI sequences, and the fusion strategy of various MRI sequences was optimized. In addition, we compared the effectiveness of the model in the parotid neoplasm classification with experienced radiologists. The deep learning model delivered reliable outcomes in differentiating benign and malignant parotid neoplasms. The model, which was trained by the fusion of T2-weighted, postcontrast T1-weighted, and diffusion-weighted imaging (b=1000 s/mm2 ), produced the best result, with an accuracy score of 0.85, an area under the receiver operator characteristic (ROC) curve of 0.96, a sensitivity score of 0.90, and a specificity score of 0.84. In addition, the multi-modal paradigm exhibited reliable outcomes in diagnosing the pleomorphic adenoma and the Warthin tumor, but not in the identification of the basal cell adenoma. An accurate and efficient AI based classification model was produced to classify parotid neoplasms, resulting from the fusion of multimodal MRI sequences. The effectiveness certainly outperformed the model with single MRI images or single MRI sequences as input, and potentially, experienced radiologists. 3 Laryngoscope, 133:327-335, 2023.

RGB Channel Superposition Algorithm with Acetowhite Mask Images in a Cervical Cancer Classification Deep Learning Model.

Cervical cancer is one of the main causes of death from cancer in women. However, it can be treated successfully at an early stage. This study aims to propose an image processing algorithm based on acetowhite, which is an important criterion for diagnosing cervical cancer, to increase the accuracy of the deep learning classification model. Then, we mainly compared the performance of the model, the original image without image processing, a mask image made with acetowhite as the region of interest, and an image using the proposed algorithm. In conclusion, the deep learning classification model based on images with the proposed algorithm achieved an accuracy of 81.31%, which is approximately 9% higher than the model with original images and approximately 4% higher than the model with acetowhite mask images. Our study suggests that the proposed algorithm based on acetowhite could have a better performance than other image processing algorithms for classifying stages of cervical images.

Attention-based bidirectional LSTM with embedding technique for classification of COVID-19 articles

The epidemic of COVID-19 has thrown the planet into an awfully tricky situation putting a terrifying end to thousands of lives; the global health infrastructure continues to be in significant danger. Several machine learning techniques and pre-defined models have been demonstrated to accomplish the classification of COVID-19 articles. These delineate strategies to extract information from structured and unstructured data sources which form the article repository for physicians and researchers. Expanding the knowledge of diagnosis and treatment of COVID-19 virus is the key benefit of these researches. A multi-label Deep Learning classification model has been proposed here on the LitCovid dataset which is a collection of research articles on coronavirus. Relevant prior articles are explored to select appropriate network parameters that could promote the achievement of a stable Artificial Neural Network mechanism for COVID-19 virus-related challenges. We have noticed that the proposed classification model achieves accuracy and micro-F1 score of 75.95% and 85.2, respectively. The experimental result also indicates that the propound technique outperforms the surviving methods like BioBERT and Longformer.

Development and Validation of a Visually Explainable Deep Learning Model for Classification of C-shaped Canals of the Mandibular Second Molars in Periapical and Panoramic Dental Radiographs

IntroductionThe purpose of this study was to develop and validate a visually explainable deep learning model for the classification of C-shaped canals of the mandibular second molars in dental radiographs. MethodsThe periapical and panoramic images of 1000 mandibular second molars were collected from 372 patients. The diagnostic performance of the deep learning system using periapical and panoramic radiographs was investigated in respect to its ability to determine whether the second mandibular molar showed a C-shaped canal configuration. The assessment of the canal configuration of cone-beam computed tomographic volumes from 372 patients (740 mandibular second molars) was used as a gold standard. ResultsThe deep convolutional neural network algorithm model showed high accuracy in predicting the C-shaped canal variation among mandibular second molars in both periapical and panoramic images. The model demonstrated best results when using image patches including only the root portion of the tooth and when using both periapical and panoramic images for training (area under the curve [AUC] = 0.99). The model's diagnostic performance using only the root portion of the tooth (AUC: periapical = 0.98 and panoramic = 0.95) was similar to a specialist (AUC: periapical = 0.95 and panoramic = 0.96) and better than a novice general clinician (AUC: periapical = 0.89 and panoramic = 0.91). Both the specialist and general clinician showed better diagnostic performance when reading panoramic radiographs compared with periapical images. ConclusionsWith further optimization of the test data using a larger data set and improvements made in the model, a deep learning system may be expected to effectively diagnose C-shaped canals and aid clinicians in practice and education.

3D Kinect Camera Scheme with Time-Series Deep-Learning Algorithms for Classification and Prediction of Lung Tumor Motility.

This paper proposes a time-series deep-learning 3D Kinect camera scheme to classify the respiratory phases with a lung tumor and predict the lung tumor displacement. Specifically, the proposed scheme is driven by two time-series deep-learning algorithmic models: the respiratory-phase classification model and the regression-based prediction model. To assess the performance of the proposed scheme, the classification and prediction models were tested with four categories of datasets: patient-based datasets with regular and irregular breathing patterns; and pseudopatient-based datasets with regular and irregular breathing patterns. In this study, ‘pseudopatients’ refer to a dynamic thorax phantom with a lung tumor programmed with varying breathing patterns and breaths per minute. The total accuracy of the respiratory-phase classification model was 100%, 100%, 100%, and 92.44% for the four dataset categories, with a corresponding mean squared error (MSE), mean absolute error (MAE), and coefficient of determination (R2) of 1.2–1.6%, 0.65–0.8%, and 0.97–0.98, respectively. The results demonstrate that the time-series deep-learning classification and regression-based prediction models can classify the respiratory phases and predict the lung tumor displacement with high accuracy. Essentially, the novelty of this research lies in the use of a low-cost 3D Kinect camera with time-series deep-learning algorithms in the medical field to efficiently classify the respiratory phase and predict the lung tumor displacement.

Deep Learning Classification Model for English Translation Styles Introducing Attention Mechanism

Both short-distance association knowledge and long-distance interaction knowledge in the knowledge base contain rich semantics. When learning entity and relation representation in a knowledge base, if we can learn short-distance association knowledge and long-distance interaction knowledge at the same time, we can learn the representation method with rich semantic information and keep the original structure of the knowledge base. Among the knowledge contained in a large number of records, some knowledge reflects individual characteristics and can be called local knowledge; others reflect group characteristics and can be called global knowledge. Using different ways to learn local and global knowledge in the deep learning model will better reflect the difference between the two kinds of knowledge at the model level, and make the model have the ability to understand both individual characteristics and overall characteristics. Through layer-by-layer forward propagation and error back propagation algorithms, the entire network is gradually optimized in an “end-to-end” manner. This “end-to-end” approach leaves some means of introducing prior knowledge flexibly into the model. Although it can reduce the burden on researchers, this “data-driven” approach brings the shortcomings of poor interpretability of learning results and weak generalization ability. Combining the specific prior knowledge implicit in the data with the deep learning algorithm can optimize the algorithm in a targeted manner and avoid blind searching in the solution space, so as to obtain a model with better performance and wider use. To this end, this paper investigates combining prior knowledge with deep learning to design efficient algorithms to address the classification of English translation styles. This paper combines local knowledge with global knowledge and deep learning methods and proposes a memory neural network method combining local knowledge and global knowledge. By recording the local knowledge in the local memory module and simultaneously recording the global knowledge in the global memory module, the method effectively learns the latent information in a large number of records. This paper combines short-distance association knowledge with long-distance interaction knowledge and a distributed representation learning method based on deep learning and proposes a deep learning method combining short-distance association knowledge and long-distance interaction knowledge. On the IWSLT English translation task, experiments show that the method significantly improves translation quality, confirming that grammatical dependencies enhance attention by supplementing dependent grammatical information, resulting in more effective and richer context vectors that more accurately represent contextual situations. Additional experimental analysis showed that the model underwent careful parameter selection and analysis. By mining valuable long-distance interactive knowledge in the knowledge base and using it in the distributed representation learning of the knowledge base, while constraining the short-distance related knowledge and constraining the long-distance interactive knowledge, the learned knowledge can be used to effectively complete the knowledge base distributed representation for discovering new relations.

Dynamic defense against byzantine poisoning attacks in federated learning

Federated learning, as a distributed learning that conducts the training on the local devices without accessing to the training data, is vulnerable to Byzantine poisoning adversarial attacks. We argue that the federated learning model has to avoid those kind of adversarial attacks through filtering out the adversarial clients by means of the federated aggregation operator. We propose a dynamic federated aggregation operator that dynamically discards those adversarial clients and allows to prevent the corruption of the global learning model. We assess it as a defense against adversarial attacks deploying a deep learning classification model in a federated learning setting on the Fed-EMNIST Digits, Fashion MNIST and CIFAR-10 image datasets. The results show that the dynamic selection of the clients to aggregate enhances the performance of the global learning model and discards the adversarial and poor (with low quality models) clients.

RiceBioS: Identification of Biotic Stress in Rice Crops Using Edge-as-a-Service

The identification of biotic stress of rice crops using handheld sensing devices is a challenge, as computationally intensive machine learning models are difficult to be executed in these devices. This challenge is exacerbated in farmers’ fields located in remote regions with limited internet connectivity. Thus, an individualiased plant-specific solution to detect biotic stress due to crop infections is required in farms adopting digital agricultural practices. The existing biotic stress detection solutions are deficient in their ability to make decisions in real-time. It is required to have a system that is capable of making decisions at the edge in handheld devices having limited computational capability. This paper proposes RiceBioS, an AI-based deep learning-enabled handheld device for identifying biotic stress in rice crops using the computational capabilities of handheld devices. RiceBioS adopts Edge-as-a-Service (EaaS) as an approach for classifying rice crop images into two categories – healthy and stressed. The biotic stress condition is further diagnosed into two types of infections, fungal (rice blast) and bacterial (bacterial leaf blight of rice) by pruning the shrunk deep learning classification model and incorporating an automated RoI detection and feature extraction workflow, which makes use of adaptive thresholding and hierarchial masking techniques to perform dimensionality reduction. While RiceBioS demonstrates a test accuracy of 93.25%, it exhibits a negligible tradeoff on a smartphone after deployment. This cutting edge solution helps the farmers make informed decisions based on real-time insights provided by the user-friendly mobile application interface of RiceBioS.

An Algorithm for Automatic Rib Fracture Recognition Combined with nnU-Net and DenseNet.

Rib fracture is the most common thoracic clinical trauma. Most patients have multiple different types of rib fracture regions, so accurate and rapid identification of all trauma regions is crucial for the treatment of rib fracture patients. In this study, a two-stage rib fracture recognition model based on nnU-Net is proposed. First, a deep learning segmentation model is trained to generate candidate rib fracture regions, and then, a deep learning classification model is trained in the second stage to classify the segmented local fracture regions according to the candidate fracture regions generated in the first stage to determine whether they are fractures or not. The results show that the two-stage deep learning model proposed in this study improves the accuracy of rib fracture recognition and reduces the false-positive and false-negative rates of rib fracture detection, which can better assist doctors in fracture region recognition.

A Cloud-Based Mapping Approach Using Deep Learning and Very-High Spatial Resolution Earth Observation Data to Facilitate the SDG 11.7.1 Indicator Computation

As urbanized areas continue to expand rapidly across all continents, the United Nations adopted in 2015 the Sustainable Development Goal (SDG) 11, aimed at shaping a sustainable future for city dwellers. Earth Observation (EO) satellite data can provide at a fine scale, essential urban land use information for computing SDG 11 indicators in order to complement or even replace inaccurate or invalid existing spatial datasets. This study proposes an EO-based approach for extracting large scale information regarding urban open spaces (UOS) and land allocated to streets (LAS) at the city level, for calculating SDG indicator 11.7.1. The research workflow was developed over the Athens metropolitan area in Greece using deep learning classification models for processing PlanetScope and Sentinel-1 imagery, employing freely-available cloud environments offered by Google. The LAS model exhibited satisfactory results while the best experiment performance for mapping UOS, considering both PlanetScope and Sentinel-1 data, yielded high commission errors, however, the cross-validation analysis with the UOS area of OpenStreetMap exhibited a total overlap of 67.38%, suggesting that our workflow is suitable for creating a “potential” UOS layer. The methodology developed herein can serve as a roadmap for the calculation of indicator 11.7.1 through national statistical offices when spatial data are absent or unreliable.

Handling partially labeled network data: A semi-supervised approach using stacked sparse autoencoder

Network traffic analytics has become a crucial task in order to better understand and manage network resources, especially in the network softwarization era where the implementation of this concept can be done easily with network function virtualization. Currently, many approaches have been proposed to improve the performance of traffic classification. However, as new types of traffic emerge every day (and they are generally not labeled), this opens a new challenge to be handled. Moreover, the question of how to accurately classify the traffic using a limited amount of labeled data or partially labeled data is also another important concern. In fact, labeling data is often difficult and time-consuming. In order to solve the previously described issues, we reformulate traffic classification into a semi-supervised learning where both supervised learning (using labeled data) and unsupervised learning (no label data) are combined. To do so, this paper presents a stacked sparse autoencoder (SSAE) based semi-supervised deep-learning model for traffic classification. The main motivations of this approach are: (i) unlabeled data is often abundant and easily available; (ii) classification performance of the whole model can be greatly improved when a large amount of unlabeled traffic is included in the training process; (iii) there is a limit to how much human effort can be thrown at the labeling problem. To investigate the performance of our approach, an empirical study has been conducted on a real dataset and results indicate that using a large amount of unlabeled data in the SSAE pre-trained phase can improve significantly the classification performance of the whole model. Furthermore, the proposed approach is compared against other representative machine-learning and deep-learning models, which are Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Multi-Layer Perceptron (MLP), eXtreme Gradient Boosting (XGBoost), and Autoencoder.

U-shaped deep-learning models for island ecosystem type classification, a case study in Con Dao Island of Vietnam

The monitoring of ecosystem dynamics utilises time and resources from scientists and land-use managers, especially in wetland ecosystems in islands that have been affected significantly by both the current state of oceans and human-made activities. Deep-learning models for natural and anthropogenic ecosystem type classification, based on remote sensing data, have become a tool to potentially replace manual image interpretation. This study proposes a U-Net model to develop a deep learning model for classifying 10 island ecosystems with cloud- and shadow-based data using Sentinel-2, ALOS and NOAA remote sensing data. We tested and compared different optimiser methods with two benchmark methods, including support vector machines and random forests. In total, 48 U-Net models were trained and compared. The U-Net model with the Adadelta optimiser and 64 filters showed the best result, because it could classify all island ecosystems with 93 percent accuracy and a loss function value of 0.17. The model was used to classify and successfully manage ecosystems on a particular island in Vietnam. Compared to island ecosystems, it is not easy to detect coral reefs due to seasonal ocean currents. However, the trained deep-learning models proved to have high performances compared to the two traditional methods. The best U-Net model, which needs about two minutes to create a new classification, could become a suitable tool for island research and management in the future.

Analyzing RNA-Seq Gene Expression Data Using Deep Learning Approaches for Cancer Classification

Ribonucleic acid Sequencing (RNA-Seq) analysis is particularly useful for obtaining insights into differentially expressed genes. However, it is challenging because of its high-dimensional data. Such analysis is a tool with which to find underlying patterns in data, e.g., for cancer specific biomarkers. In the past, analyses were performed on RNA-Seq data pertaining to the same cancer class as positive and negative samples, i.e., without samples of other cancer types. To perform multiple cancer type classification and to find differentially expressed genes, data for multiple cancer types need to be analyzed. Several repositories offer RNA-Seq data for various cancer types. In this paper, data from the Mendeley data repository for five cancer types are analyzed. As a first step, RNA-Seq values are converted to 2D images using normalization and zero padding. In the next step, relevant features are extracted and selected using Deep Learning (DL). In the last phase, classification is performed, and eight DL algorithms are used. Results and discussion are based on four different splitting strategies and k-fold cross validation for each DL classifier. Furthermore, a comparative analysis is performed with state of the art techniques discussed in literature. The results demonstrated that classifiers performed best at 70–30 split, and that Convolutional Neural Network (CNN) achieved the best overall results. Hence, CNN is the best DL model for classification among the eight studied DL models, and is easy to implement and simple to understand.

Internet of things and multi-class deep feature-fusion based classification of tomato leaf disease

A deep transfer learning (deep-TL) classification model has been proposed to diagnose tomato leaf disease. The main challenge of inaccurate classification of a convolution neural network (CNN) model was the availability of the small-sized dataset. This model deals with the challenges like availability of small-sized and imbalanced datasets. The proposed Alex support vector machine (SVM) fused hybrid classification (ASFHC) model is based on fully fusion technology that avoids overfitting to classify the type of disease in tomato leaves. The proposed model achieves the best performance in terms of accuracy by data augmentation of the training data. It uses a pre-trained network for feature extraction with the modification of architecture by concatenating two layers FC6 and FC7 (fully connected layer), plus a linear SVM classifier for classification of the disease. The uniqueness of the research is although the dataset is not balanced, the performance of the model has achieved the maximum. Compared with VGG 16 and VGG 19, the proposed model (ASFHC) has been evaluated using different measuring parameters, indicating remarkable computation time for implementation in the internet of things (IoT) domain. The overall accuracy attained by the model is 99.62%.