Deep Learning-based Classification Methods Research Articles

Deep Learning-Based Classification Methods for Detection of Diseases in Rice Leaves – A Review

Deep Learning-Based Classification Methods for Detection of Diseases in Rice Leaves – A Review

Using deep learning-based classification methods for interpreting brain MRI images for tumor diagnosis

Using deep learning-based classification methods for interpreting brain MRI images for tumor diagnosis

MycoAI: Fast and accurate taxonomic classification for fungal ITS sequences.

Efficient and accurate classification of DNA barcode data is crucial for large-scale fungal biodiversity studies. However, existing methods are either computationally expensive or lack accuracy. Previous research has demonstrated the potential of deep learning in this domain, successfully training neural networks for biological sequence classification. We introduce the MycoAI Python package, featuring various deep learning models such as BERT and CNN tailored for fungal Internal Transcribed Spacer (ITS) sequences. We explore different neural architecture designs and encoding methods to identify optimal models. By employing a multi-head output architecture and multi-level hierarchical label smoothing, MycoAI effectively generalizes across the taxonomic hierarchy. Using over 5 million labelled sequences from the UNITE database, we develop two models: MycoAI-BERT and MycoAI-CNN. While we emphasize the necessity of verifying classification results by AI models due to insufficient reference data, MycoAI still exhibits substantial potential. When benchmarked against existing classifiers such as DNABarcoder and RDP on two independent test sets with labels present in the training dataset, MycoAI models demonstrate high accuracy at the genus and higher taxonomic levels, with MycoAI-CNN being the fastest and most accurate. In terms of efficiency, MycoAI models can classify over 300,000 sequences within 5 min. We publicly release the MycoAI models, enabling mycologists to classify their ITS barcode data efficiently. Additionally, MycoAI serves as a platform for developing further deep learning-based classification methods. The source code for MycoAI is available under the MIT Licence at https://github.com/MycoAI/MycoAI.

CCSI: Continual Class-Specific Impression for data-free class incremental learning

In real-world clinical settings, traditional deep learning-based classification methods struggle with diagnosing newly introduced disease types because they require samples from all disease classes for offline training. Class incremental learning offers a promising solution by adapting a deep network trained on specific disease classes to handle new diseases. However, catastrophic forgetting occurs, decreasing the performance of earlier classes when adapting the model to new data. Prior proposed methodologies to overcome this require perpetual storage of previous samples, posing potential practical concerns regarding privacy and storage regulations in healthcare. To this end, we propose a novel data-free class incremental learning framework that utilizes data synthesis on learned classes instead of data storage from previous classes. Our key contributions include acquiring synthetic data known as Continual Class-Specific Impression (CCSI) for previously inaccessible trained classes and presenting a methodology to effectively utilize this data for updating networks when introducing new classes. We obtain CCSI by employing data inversion over gradients of the trained classification model on previous classes starting from the mean image of each class inspired by common landmarks shared among medical images and utilizing continual normalization layers statistics as a regularizer in this pixel-wise optimization process. Subsequently, we update the network by combining the synthesized data with new class data and incorporate several losses, including an intra-domain contrastive loss to generalize the deep network trained on the synthesized data to real data, a margin loss to increase separation among previous classes and new ones, and a cosine-normalized cross-entropy loss to alleviate the adverse effects of imbalanced distributions in training data. Extensive experiments show that the proposed framework achieves state-of-the-art performance on four of the public MedMNIST datasets and in-house echocardiography cine series, with an improvement in classification accuracy of up to 51% compared to baseline data-free methods. Our code is available at https://github.com/ubc-tea/Continual-Impression-CCSI.

Event recognition method based on feature synthesizing for a zero-shot intelligent distributed optical fiber sensor

Phase-sensitive optical time domain reflectometer (Φ-OTDR) is an emergent distributed optical sensing system with the advantages of high localization accuracy and high sensitivity. It has been widely used for intrusion identification, pipeline monitoring, under-ground tunnel monitoring, etc. Deep learning-based classification methods work well for Φ-OTDR event recognition tasks with sufficient samples. However, the lack of training data samples is sometimes a serious problem for these data-driven algorithms. This paper proposes a novel feature synthesizing approach to solve this problem. A mixed class approach and a reinforcement learning-based guided training method are proposed to realize high-quality feature synthesis. Experiment results in the task of eight event classifications, including one unknown class, show that the proposed method can achieve an average classification accuracy of 42% for the unknown class and obtain its event type, meanwhile achieving a 74% average overall classification accuracy. This is 29% and 7% higher, respectively, than those of the ordinary instance synthesizing method. Moreover, this is the first time that the Φ-OTDR system can recognize a specific event and tell its event type without collecting its data sample in advance.

Dilated Spectral–Spatial Gaussian Transformer Net for Hyperspectral Image Classification

In recent years, deep learning-based classification methods for hyperspectral images (HSIs) have gained widespread popularity in fields such as agriculture, environmental monitoring, and geological exploration. This is owing to their ability to automatically extract features and deliver outstanding performance. This study provides a new Dilated Spectral–Spatial Gaussian Transformer Net (DSSGT) model. The DSSGT model incorporates dilated convolutions as shallow feature extraction units, which allows for an expanded receptive field while maintaining computational efficiency. We integrated transformer architecture to effectively capture feature relationships and generate deep fusion features, thereby enhancing classification accuracy. We used consecutive dilated convolutional layers to extract joint low-level spectral–spatial features. We then introduced Gaussian Weighted Pixel Embedding blocks, which leverage Gaussian weight matrices to transform the joint features into pixel-level vectors. By combining the features of each pixel with its neighbouring pixels, we obtained pixel-level representations that are more expressive and context-aware. The transformed vector matrix was fed into the transformer encoder module, enabling the capture of global dependencies within the input data and generating higher-level fusion features with improved expressiveness and discriminability. We evaluated the proposed DSSGT model using five hyperspectral image datasets through comparative experiments. Our results demonstrate the superior performance of our approach compared to those of current state-of-the-art methods, providing compelling evidence of the DSSGT model’s effectiveness.

Multiscale Feature Extraction by Using Convolutional Neural Network: Extraction of Objects from Multiresolution Images of Urban Areas

The emergence of deep learning-based classification methods has led to considerable advancements and remarkable performance in image recognition. This study introduces the Multiscale Feature Convolutional Neural Network (MSFCNN) for the extraction of complex urban land cover data, with a specific emphasis on buildings and roads. MSFCNN is employed to extract multiscale features from three distinct image types—Unmanned Aerial Vehicle (UAV) images, high-resolution satellite images (HR), and low-resolution satellite images (LR)—all collected within the Fengshan District of Kaohsiung, Taiwan. The model in this study demonstrated remarkable accuracy in classifying two key land cover categories. Its success in extracting multiscale features from different image resolutions. In the case of UAV images, MSFCNN achieved an accuracy rate of 91.67%, with a Producer’s Accuracy (PA) of 93.33% and a User’s Accuracy (UA) of 90.0%. Similarly, the model exhibited strong performance with HR images, yielding accuracy, PA, and UA values of 92.5%, 93.33%, and 91.67%, respectively. These results closely align with those obtained for LR imagery, which achieved respective accuracy rates of 93.33%, 95.0%, and 91.67%. Overall, the MSFCNN excels in the classification of both UAV and satellite images, showcasing its versatility and robustness across various data sources. The model is well suited for the task of updating cartographic data related to urban buildings and roads.

Deep Learning-Based Method for Classification and Ripeness Assessment of Fruits and Vegetables

Food waste is a global concern and is the focus of this research. Currently, no method in the state of the art classifies multiple fruits and vegetables and their level of ripening. The objective of the study is to design and develop an intelligent system based on deep learning techniques to classify between types of fruits and vegetables, and also to evaluate the level of ripeness of some of them. The system consists of two models using the MobileNet V2 architecture. One algorithm is for the classification of 32 classes of fruits and vegetables, and another is for the determination of the ripeness of 6 classes of them. The overall intelligent system is the union of the two models, predicting first the class of fruit or vegetable and then its ripeness. The fruits and vegetables classification model achieved 97.86% accuracy, 98% precision, 98% recall, and 98% F1-score, while the ripeness assessment model achieved 100% accuracy, 98% precision, 99% recall, and 99% F1-score. According to the results, the proposed system is able to classify between types of fruits and vegetables and evaluate their ripeness. To achieve the best performance indicators, it is necessary to obtain the appropriate hyperparameters for the artificial intelligence models, in addition to having an extensive database with well-defined classes.

An Overview of Hyperspectral Image Classification by Data-driven Deep Learning

Hyperspectral imaging (HSI) in remote sensing is gaining significant attention due to its complexity, posing challenges for conventional machine learning in achieving accurate classification. The inherent nonlinear relationship between captured spectral information and materials further complicates hyperspectral imaging. Deep learning has emerged as an effective tool for feature extraction, finding widespread applications in image processing tasks. Motivated by its success, this survey integrates deep learning into hyperspectral imaging (HSI) classification, demonstrating commendable performance. The paper systematically reviews existing literature, providing a comparative analysis of strategies. Primary challenges in HSI classification for traditional methods are outlined, emphasizing the advantages of deep learning. Our framework categorizes works into three types: spectral-feature networks, spatial-feature networks, and spectral-spatial-feature networks, offering a comprehensive review of recent achievements and diverse approaches. Considering limited training samples in remote sensing and substantial data requirements for deep networks, strategies to enhance classification performance are presented, offering valuable insights for future studies. Experiments apply representative deep learning-based classification methods to real HSIs, providing practical validation. The survey contributes to understanding the current landscape in deep learning-based HSI classification and lays a foundation for future research in this evolving field.

Plant Image Classification with Nonlinear Motion Deblurring Based on Deep Learning

Despite the significant number of classification studies conducted using plant images, studies on nonlinear motion blur are limited. In general, motion blur results from movements of the hands of a person holding a camera for capturing plant images, or when the plant moves owing to wind while the camera is stationary. When these two cases occur simultaneously, nonlinear motion blur is highly probable. Therefore, a novel deep learning-based classification method applied on plant images with various nonlinear motion blurs is proposed. In addition, this study proposes a generative adversarial network-based method to reduce nonlinear motion blur; accordingly, the method is explored for improving classification performance. Herein, experiments are conducted using a self-collected visible light images dataset. Evidently, nonlinear motion deblurring results in a structural similarity index measure (SSIM) of 73.1 and a peak signal-to-noise ratio (PSNR) of 21.55, whereas plant classification results in a top-1 accuracy of 90.09% and F1-score of 84.84%. In addition, the experiment conducted using two types of open datasets resulted in PSNRs of 20.84 and 21.02 and SSIMs of 72.96 and 72.86, respectively. The proposed method of plant classification results in top-1 accuracies of 89.79% and 82.21% and F1-scores of 84% and 76.52%, respectively. Thus, the proposed network produces higher accuracies than the existing state-of-the-art methods.

Hygeia: A Multilabel Deep Learning-Based Classification Method for Imbalanced Electrocardiogram Data.

Electrocardiogram (ECG) is a common diagnostic indicator of heart disease in hospitals. Because of the low price and noninvasiveness of ECG diagnosis, it is widely used for prescreening and physical examination of heart diseases. In several studies on ECG analysis, only rough diagnoses are made to determine whether ECGs are abnormal or on a few kinds of ECG. In actual scenarios, doctors must analyze ECG samples in detail, which is a multilabel classification problem. Herein, we propose Hygeia, a multilabel deep learning-based ECG classification method that can analyze and classify 55 types of ECG. First, a guidance model is constructed to transform the multilabel classification problem into multiple interrelated two-classification models. This method ensures the good performance of each ECG analysis model, and the relationship between various types of ECG can be used in the analysis. The imbalance of samples in ECG datasets makes it difficult to analyze abnormal ECGs with high sensitivity and accuracy. We used data generation and mixed sampling methods for 11 ECG types with imbalanced problems to improve the average accuracy, sensitivity, F1 value, and accuracy from 87.74%, 43.11%, 0.3929, and 0.3929, to 92.68%, 96.92, 0.9287, and 99.47%, respectively. The average accuracy, sensitivity, F1 value, and accuracy of 44 of the 55 tags of the abnormal ECG analysis model are 99.69%, 95.81%, 0.9758, and 99.72%, respectively.

Tensor-RT-Based Transfer Learning Model for Lung Cancer Classification.

Cancer is a leading cause of death across the globe, in which lung cancer constitutes the maximum mortality rate. Early diagnosis through computed tomography scan imaging helps to identify the stages of lung cancer. Several deep learning-based classification methods have been employed for developing automatic systems for the diagnosis and detection of computed tomography scan lung slices. However, the diagnosis based on nodule detection is a challenging task as it requires manual annotation of nodule regions. Also, these computer-aided systems have yet not achieved the desired performance in real-time lung cancer classification. In the present paper, a high-speed real-time transfer learning-based framework is proposed for the classification of computed tomography lung cancer slices into benign and malignant. The proposed framework comprises of three modules: (i) pre-processing and segmentation of lung images using K-means clustering based on cosine distance and morphological operations; (ii) tuning and regularization of the proposed model named as weighted VGG deep network (WVDN); (iii) model inference in Nvidia tensor-RT during post-processing for the deployment in real-time applications. In this study, two pre-trained CNN models were experimented and compared with the proposed model. All the models have been trained on 19,419 computed tomography scan lung slices, which were obtained from the publicly available Lung Image Database Consortium and Image Database Resource Initiative dataset. The proposed model achieved the best classification metric, an accuracy of 0.932, precision, recall, an F1 score of 0.93, and Cohen's kappa score of 0.85. A statistical evaluation has also been performed on the classification parameters and achieved a p-value <0.0001 for the proposed model. The quantitative and statistical results validate the improved performance of the proposed model as compared to state-of-the-art methods. The proposed framework is based on complete computed tomography slices rather than the marked annotations and may help in improving clinical diagnosis.

A survey of deep learning-based classification methods for steady-state visual evoked potentials

Purpose Steady-state visual evoked potential (SSVEP) based BCI has attracted great interests owing to the high information transfer rate (ITR) and little training requirement. The performance of SSVEP-based BCI heavily depends on the classification methods. Deep Learning (DL) technology provides an alternative avenue for the data classification in SSVEP-based BCI, and has received increasing interests in recent years. This review aimed to summarize the progress of DL-based classification methods for SSVEP data over the past decade.Materials and method The literature was searched and selected based on the research topics of DL and SSVEP. We categorized these methods into four classes, i.e., traditional neural network structures-based DL methods, traditional frequency recognition methods inspiring DL methods, attention mechanisms-based DL models, and transfer learning technology-based DL methods, and generative model-based recognition method. Moreover, we analyzed the current challenges and presented future research opportunities.Conclusions This study provides a systematic description of the current development status on DL-based SSVEP classification methods, and sheds insight on future researches.

A Bayesian Deep Learning Approach With Convolutional Feature Engineering to Discriminate Cyber-Physical Intrusions in Smart Grid Systems

The emergence of cyber-physical smart grid (CPSG) systems has revolutionized the traditional power grid by enabling the bidirectional energy flow between consumers and utilities. However, due to escalated information exchange between the end-users, it has posed a greater challenge to the cyber security mechanisms for the communication networks at the cyber and physical planes. To address these challenges, we propose a Bayesian approach integrated with deep convolutional neural networks (CNN-Bayesian). While, the Bayesian component is used to discriminate cyber-physical intrusions from the normal events in the binary and multi-class events. CNN layers are utilized to handle the high-dimensional feature space prior to the intrusions classification task. The proposed method is validated using real-time Industrial control systems (ICS) dataset against the standard deep learning-based classification methods such as recurrent neural networks (RNN) and long-short term memory (LSTM). From the experimental results, it can be inferred that the proposed CNN-Bayesian method outperforms the existing benchmark classification methods to discriminate intrusions in CPSG systems using evaluation metrics such as accuracy, precision, recall, and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula> -score.

Deep Learning-Based Method for Classification of Sugarcane Varieties

The classification of sugarcane varieties using products derived from remote sensing allows for the monitoring of plants with different profiles without necessarily having physical contact with the study objects. However, differentiating between varieties can be challenging due to the similarity of the spectral characteristics of each crop. Thus, this study aimed to classify four sugarcane varieties through deep neural networks, subsequently comparing the results with traditional machine learning techniques. In order to provide more data as input for the classification models, along with the multi-band values of the pixels and vegetation indices, other information can be obtained from the sensor bands through RGB combinations by reconciling different bands so as to yield the characteristics of crop varieties. The methodology created to discriminate sugarcane varieties consisted of a dense neural network, with the number of hidden layers determined by the greedy layer-wise method and multiples of four neurons in each layer; additionally, a 5-fold evaluation in the training data was composed of Sentinel-2 band data, vegetation indices, and RGB combinations. Comparing the results acquired from each model with the hyperparameters selected by Bayesian optimisation, except for the neural network with manually defined parameters, it was possible to observe a greater precision of 99.55% in the SVM model, followed by the neural network developed by the study, random forests, and kNN. However, the final neural network model prediction resulted in the 99.48% accuracy of a six-hidden-layers network, demonstrating the potential of using neural networks in classification. Among the characteristics that contributed the most to the classification, the chlorophyll-sensitive bands, especially B6, B7, B11, and some RGB combinations, had the most impact on the correct classification of samples by the neural network model. Thus, the regions encompassing the near-infrared and shortwave infrared regions proved to be suitable for the discrimination of sugarcane varieties.

Deep Learning and Entropy-Based Texture Features for Color Image Classification.

In the domain of computer vision, entropy-defined as a measure of irregularity-has been proposed as an effective method for analyzing the texture of images. Several studies have shown that, with specific parameter tuning, entropy-based approaches achieve high accuracy in terms of classification results for texture images, when associated with machine learning classifiers. However, few entropy measures have been extended to studying color images. Moreover, the literature is missing comparative analyses of entropy-based and modern deep learning-based classification methods for RGB color images. In order to address this matter, we first propose a new entropy-based measure for RGB images based on a multivariate approach. This multivariate approach is a bi-dimensional extension of the methods that have been successfully applied to multivariate signals (unidimensional data). Then, we compare the classification results of this new approach with those obtained from several deep learning methods. The entropy-based method for RGB image classification that we propose leads to promising results. In future studies, the measure could be extended to study other color spaces as well.

Integration of Object-Based Image Analysis and Convolutional Neural Network for the Classification of High-Resolution Satellite Image: A Comparative Assessment

During the past decade, deep learning-based classification methods (e.g., convolutional neural networks—CNN) have demonstrated great success in a variety of vision tasks, including satellite image classification. Deep learning methods, on the other hand, do not preserve the precise edges of the targets of interest and do not extract geometric features such as shape and area. Previous research has attempted to address such issues by combining deep learning with methods such as object-based image analysis (OBIA). Nonetheless, the question of how to integrate those methods into a single framework in such a way that the benefits of each method complement each other remains. To that end, this study compared four integration frameworks in terms of accuracy, namely OBIA artificial neural network (OBIA ANN), feature fusion, decision fusion, and patch filtering, according to the results. Patch filtering achieved 0.917 OA, whereas decision fusion and feature fusion achieved 0.862 OA and 0.860 OA, respectively. The integration of CNN and OBIA can improve classification accuracy; however, the integration framework plays a significant role in this. Future research should focus on optimizing the existing CNN and OBIA frameworks in terms of architecture, as well as investigate how CNN models should use OBIA outputs for feature extraction and classification of remotely sensed images.

A deep learning-based method for classification, detection, and localization of weeds in turfgrass.

Precision spraying of synthetic herbicides can reduce herbicide input. Previous research demonstrated the effectiveness of using image classification neural networks for detecting weeds growing in turfgrass, but did not attempt to discriminate weed species and locate the weeds on the input images. The objectives of this research were to: (i) investigate the feasibility of training deep learning models using grid cells (subimages) to detect the location of weeds on the image by identifying whether or not the grid cells contain weeds; and (ii) evaluate DenseNet, EfficientNetV2, ResNet, RegNet and VGGNet to detect and discriminate multiple weed species growing in turfgrass (multi-classifier) and detect and discriminate weeds (regardless of weed species) and turfgrass (two-classifier). The VGGNet multi-classifier exhibited an F1 score of 0.950 when used to detect common dandelion and achieved high F1 scores of ≥0.983 to detect and discriminate the subimages containing dallisgrass, purple nutsedge and white clover growing in bermudagrass turf. DenseNet, EfficientNetV2 and RegNet multi-classifiers exhibited high F1 scores of ≥0.984 for detecting dallisgrass and purple nutsedge. Among the evaluated neural networks, EfficientNetV2 two-classifier exhibited the highest F1 scores (≥0.981) for exclusively detecting and discriminating subimages containing weeds and turfgrass. The proposed method can accurately identify the grid cells containing weeds and thus precisely locate the weeds on the input images. Overall, we conclude that the proposed method can be used in the machine vision subsystem of smart sprayers to locate weeds and make the decision for precision spraying herbicides onto the individual map cells. © 2022 Society of Chemical Industry.

Deep learning-based classification of the anterior chamber angle in glaucoma gonioscopy.

In the proposed network, the features were first extracted from the gonioscopically obtained anterior segment photographs using the densely-connected high-resolution network. Then the useful information is further strengthened using the hybrid attention module to improve the classification accuracy. Between October 30, 2020, and January 30, 2021, a total of 146 participants underwent glaucoma screening. One thousand seven hundred eighty original images of the ACA were obtained with the gonioscope and slit lamp microscope. After data augmentation, 4457 images are used for the training and validation of the HahrNet, and 497 images are used to evaluate our algorithm. Experimental results demonstrate that the proposed HahrNet exhibits a good performance of 96.2% accuracy, 99.0% specificity, 96.4% sensitivity, and 0.996 area under the curve (AUC) in classifying the ACA test dataset. Compared with several deep learning-based classification methods and nine human readers of different levels, the HahrNet achieves better or more competitive performance in terms of accuracy, specificity, and sensitivity. Indeed, the proposed ACA classification method will provide an automatic and accurate technology for the grading of glaucoma.

Deep and Lightweight Neural Network for Histopathological Image Classification

Breast cancer is a fatal disease affecting women, and early detection and proper treatment are crucial. Classifying medical images correctly is the first and most important step in the cancer diagnosis stage. Deep learning-based classification methods in various domains demonstrate advances in accuracy. However, as deep learning improves, the layers of neural networks get deeper, raising challenges, such as overfitting and gradient vanishing. For instance, a medical image is simpler than an ordinary one, making it vulnerable to overfitting issues. We present breast histopathological classification methods with two deep neural networks, Xception and LightXception with aid of voting schemes over split images. Most deep neural networks classify thousands classes of images, but the breast histopathological image classes are far fewer than those of other image classification tasks. Because the BreakHis dataset is relatively simpler than typical image datasets, such as ImageNet, applying the conventional highly deep neural networks may suffer from the aforementioned overfitting or gradient vanishing problems. Additionally, highly deep neural networks require more resources, leading to high computational costs. Consequently, we propose a new network; LightXception by cutting off layers at the bottom of the Xception network and reducing the number of channels of convolution filters. LightXception has only about 35% of parameters compared to those of the original Xception network with minimal expense on performance. Based on images with 100X magnification factor, the performance comparisons for Xception vs. LightXception are 97.42% vs. 97.31% on classification accuracy, 97.42% vs. 97.42% on recall, and 99.26% vs. 98.67% of precision.