Classification Error Rate Research Articles

Optimizing the performance of convolutional neural network for enhanced gesture recognition using sEMG

Deep neural networks (DNNs) have demonstrated higher performance results when compared to traditional approaches for implementing robust myoelectric control (MEC) systems. However, the delay induced by optimising a MEC remains a concern for real-time applications. As a result, an optimised DNN architecture based on fine-tuned hyperparameters is required. This study investigates the optimal configuration of convolutional neural network (CNN)-based MEC by proposing an effective data segmentation technique and a generalised set of hyperparameters. Firstly, two segmentation strategies (disjoint and overlap) and various segment and overlap sizes were studied to optimise segmentation parameters. Secondly, to address the challenge of optimising the hyperparameters of a DNN-based MEC system, the problem has been abstracted as an optimisation problem, and Bayesian optimisation has been used to solve it. From 20 healthy people, ten surface electromyography (sEMG) grasping movements abstracted from daily life were chosen as the target gesture set. With an ideal segment size of 200 ms and an overlap size of 80%, the results show that the overlap segmentation technique outperforms the disjoint segmentation technique (p-value < 0.05). In comparison to manual (12.76 ± 4.66), grid (0.10 ± 0.03), and random (0.12 ± 0.05) search hyperparameters optimisation strategies, the proposed optimisation technique resulted in a mean classification error rate (CER) of 0.08 ± 0.03 across all subjects. In addition, a generalised CNN architecture with an optimal set of hyperparameters is proposed. When tested separately on all individuals, the single generalised CNN architecture produced an overall CER of 0.09 ± 0.03. This study's significance lies in its contribution to the field of EMG signal processing by demonstrating the superiority of the overlap segmentation technique, optimizing CNN hyperparameters through Bayesian optimization, and offering practical insights for improving prosthetic control and human–computer interfaces.

A hyperspectral band selection algorithm for identifying high oleic acid peanuts

High oleic acid peanuts have higher oleic acid content and stronger oxidation stability than common peanuts, but their appearances are similar, which imposes difficulties for classifying. Based on this, the study aims to classify high oleic acid peanut to ensure its purity by using hyperspectral imaging technology. However, classification accuracy and efficiency are limited given the large amount of redundant information of hyperspectral images. The band iteration algorithm (BIA) is proposed to select characteristic bands by reducing the redundant information between spectral bands for the peanut classification. Hyperspectral images with 616 bands (from 400 nm to 1100 nm) of 126 high oleic acid peanuts and 126 common peanuts were collected. Then, BIA selected optimal bands as characteristic bands from adjacent bands according to the classification accuracy of each band subsets. Thirdly, three classification models, namely linear discriminant analysis, support vector machine, and partial least squares-discriminant analysis (PLS-DA), were employed to compare the performance of BIA with successive projections algorithm and competitive adaptive reweighted sampling, respectively. The experimental results show that BIA can effectively improve the classification ability of spectral data. The BIA-PLS-DA model had the best classification efficiency, and the accuracy of the test set reached 93.26%. For peanut individuals, only one peanut sample was misclassified with a classification error rate of 1.43%.

Enhancing blood flow prediction in multi-exposure laser speckle contrast imaging through ensemble learning with K-mean clustering

Purpose. Accurately visualizing and measuring blood flow is of utmost importance in maintaining optimal health and preventing the onset of various chronic diseases. One promising imaging technique that aids in visualizing perfusion in biological tissues is Multi-exposure Laser Speckle Contrast Imaging (MELSCI). MELSCI technique allows real-time quantitative measurements using multiple exposure times to obtain precise and reliable blood flow data. Additionally, the application of machine learning (ML) techniques can further enhance the accuracy of blood flow prediction in this imaging modality. Method. Our study focused on developing and evaluating Ensemble Learning ML techniques along with clustering algorithms for predicting blood flow rates in MELSCI. The effectiveness of these techniques was assessed using performance parameters, including accuracy, F1-score, precision, recall, specificity, and classification error rate. Result. Notably, the study revealed that Ensemble Learning with clustering emerged as the most accurate technique, achieving an impressive accuracy rate of 98.5%. Furthermore, it demonstrated a high recall of more than 91%, F1-score, the precision of more than 90%, higher specificity of 99%, and least classification error of 1.5%, highlighting its suitability and sustainability for flow prediction in MELSCI. Conclusion. The study’s findings imply that Ensemble Learning can significantly contribute to enhancing the accuracy of blood flow prediction in MELSCI. This advancement holds substantial promise for healthcare professionals and researchers, as it facilitates improved understanding and assessment of perfusion within biological tissues, which will contribute to the maintenance of good health and prevention of chronic diseases.

Micro aneurysm detection using optimized residual-based temporal attention Convolutional Neural Network with Inception-V3 transfer learning.

In this manuscript micro aneurysm detection using residual-based temporal attention Convolutional Neural Network (CNN) with Inception-V3 transfer learning optimized with equilibrium optimization algorithm (MA-RTCNN-Inception V3-EOA) is proposed. The proposed research work contains four phases: (1) pre-processing, (2) segmentation, (3) post-processing, and (4) classification. At first, guided box filtering for contrast enhancement and background exclusion of input image. The proposed MA-RTCNN-Inception V3-EOA based classification framework is implemented in MATLAB using several performances evaluating metrics like precision, sensitivity, f-measure, specificity, accuracy, classification error rate, and Matthews's correlation coefficient and RoC analysis. The experimental outcome demonstrates that the proposed method provides 23.56%, 14.99%, and 21.37% higher accuracy and 31.26%, 57.69%, and 21.14% minimum classification error rate compared to existing methods, such as diabetic retinopathy identification utilizing prognosis of micro aneurysm and early diagnosis for non-proliferative diabetic retinopathy depending on deep learning approaches (DRD-CNN-NPDR), a magnified adaptive feature pyramid network for automatic micro aneurysms identification (MAFPN-AMD-MAFP-Net) respectively. RESEARCH HIGHLIGHTS: Micro aneurysm detection using residual-based temporal attention Convolutional Neural Network (CNN) is proposed. To get rid of the retina background, guided box filtering is applied. COAT is used for segmenting the images into smaller parts RTCNN is used for accurate micro aneurysms disease classification. RT-CNN algorithm successfully identifies the micro aneurysms using EOA.

A Classification Approach for University English Teaching Resources Based on Multi-Feature Fusion

Abstract This study focuses on the problem of efficient classification of university English teaching resources, aiming to improve the ease of browsing and utilization of resources. To address the challenges of existing resource classification, the study proposes a label generation classification algorithm with multi-feature fusion. The algorithm incorporates TF-IDF and location information weights based on TextRank to generate labels containing corpus and location information. The performance of the algorithm is tested and simulation experiments are conducted. The study results show that the mean value of the label generation classification algorithm after multi-feature fusion reaches 0.234, 0.6219, and 0.3632 in terms of accuracy, recall, and F-value, respectively. In the simulation experiments, the algorithm achieves a classification speed of 90.33% faster than that of the traditional single-feature classification algorithm, and the classification error rate is no more than 10%. The label generation classification algorithm has a significant advantage in overall performance and effectively solves the problem of classifying university English teaching resources. This finding is of great significance for improving the utilization of teaching resources and optimizing the management of teaching resources.

An Interpretable Siamese Attention Res‐CNN for Fingerprint Spoofing Detection

In recent years, fingerprint authentication has gained widespread adoption in diverse identification systems, including smartphones, wearable devices, and attendance machines, etc. Nonetheless, these systems are vulnerable to spoofing attacks from suspicious fingerprints, posing significant risks to privacy. Consequently, a fingerprint presentation attack detection (PAD) strategy is proposed to ensure the security of these systems. Most of the previous work concentrated on how to build a deep learning framework to improve the PAD performance by augmenting fingerprint samples, and little attention has been paid to the fundamental difference between live and fake fingerprints to optimize feature extractors. This paper proposes a new fingerprint liveness detection method based on Siamese attention residual convolutional neural network (Res‐CNN) that offers an interpretative perspective to this challenge. To leverage the variance in ridge continuity features (RCFs) between live and fake fingerprints, a Gabor filter is utilized to enhance the texture details of the fingerprint ridges, followed by the construction of an attention Res‐CNN model to extract RCF between the live and fake fingerprints. The model mitigates the performance deterioration caused by gradient disappearance. Furthermore, to highlight the difference in RCF, a Siamese attention residual network is devised, and the ridge continuity amplification loss function is designed to optimize the training process. Ultimately, the RCF parameters are transferred to the model, and transfer learning is utilized to aid its acquisition, thereby assuring the model’s interpretability. The experimental outcomes conducted on three publicly accessible fingerprint datasets demonstrate the superiority of the proposed method, exhibiting remarkable performance in both true detection rate and average classification error rate. Moreover, our method exhibits remarkable capabilities in PAD tasks, including cross‐material experiments and cross‐sensor experiments. Additionally, we leverage Gradient‐weighted Class Activation Mapping to generate a heatmap that visualizes the interpretability of our model, offering a compelling visual validation.

Design a 1D-CNN'S to classify surface EMG (SEMG) signals

Amputate forearm, finger, or hand is a biggest problem for the disabled subject. Therefore, Prosthetic does a significant role for the amputees to modify the capability and mobility of their systematic activities. Using the EMG signals of hand and finger motion discrimination are continuously growth for numerous hand and finger gestures. The main problem in designing a prosthetic hand is the classification of EMG signals. Machine learning (ML) algorithms present a solution to this problem by providing a way to classify EMG signals with simply and less costly scheme. This study presents more than one experiment on two datasets in order to classify individual fingers (IF) with wrist and victory based on a normative dataset of EMG signals and Deep Learning DL. These experiments show that the overall performance (average accuracy) of the proposed method is 98.83% and the overall error classification rate (error rate) is 1.17%.

Is The Wealth Index Better than The Proxy Means Test in Poverty Targeting? A Study in Brebes and East Jakarta

The ranking of household welfare status in targeting recipients of social protection programs is important and needs attention. Appropriate welfare status ranking is one of the keys for making the various types of programs designed by the government right on target. The Proxy Means Test method is popular in Indonesia in the 2015 Integrated Database Updating. Based on another popular statistical approach to ranking welfare status, the Wealth Index method is also known. Global surveys, such as Demographic and Health Surveys, Multiple Indicator Cluster Surveys, and World Food Program Surveys, have always used the Wealth Index to rank household welfare. Using Susenas data from March 2017 to March 2022, this study found that the Proxy Means Test method is better than the Wealth Index method in both Brebes Regency and East Jakarta City. The value of the classification error rate in Brebes Regency and East Jakarta City using the Proxy Means Test method is 13.94 percent and 10.37 percent, respectively. In comparison, the Wealth Index method is 25.12 percent and 14.74 percent. This research emphasizes that the results of the ranking of household welfare status are not only influenced by the method used but also by the socioeconomic conditions and characteristics of households data in the areas targeted by the program.

Classification of colon cancer patients into consensus molecular subtypes using support vector machines.

The molecular heterogeneity of colon cancer has made classification of tumors a requirement for effective treatment. One of the approaches for molecular subtyping of colon cancer patients is the consensus molecular subtypes (CMS), developed by the Colorectal Cancer Subtyping Consortium. CMS-specific RNA-Seq-dependent classification approaches are recent, with relatively low sensitivity and specificity. In this study, we aimed to classify patients into CMS groups using their RNA-seq profiles. We first identified subtype-specific and survival-associated genes using the Fuzzy C-Means algorithm and log-rank test. We then classified patients using support vector machines with backward elimination methodology. We optimized RNA-seq-based classification using 25 genes with a minimum classification error rate. In this study, we reported the classification performance using precision, sensitivity, specificity, false discovery rate, and balanced accuracy metrics. We present a gene list for colon cancer classification with minimum classification error rates and observed the lowest sensitivity but the highest specificity with CMS3-associated genes, which significantly differed due to the low number of patients in the clinic for this group.

Smartphone application assisted optoelectronic nose for monitoring milk freshness

This study introduces an innovative method for detecting the freshness of milk through the use of smart packaging and a smartphone application. The proposed technique employs a colorimetric sensor array and a decision tree classifier, with a curvature test for feature selection. The freshness of milk samples was assessed using six colorimetric sensors, of which three, namely (Methyl violet, Chlorophenol red, and Nile blue) were chosen based on their classification error rates of 4%, 35.67%, and 8.33%, respectively. Sensor 1 was ultimately selected due to its superior performance, and its color information, along with the decision tree classification rules, was incorporated into a smartphone application. The application underwent an evaluation process, during which it was tested on twenty milk samples until total bacterial counts (TBC) reach 7 log CFU.g−1. The results showed that the application achieved a zero-error rate in identifying each sample, with the exception of 4–4.5 log CFU.g−1, where an error rate of 35% was observed. Overall, the total error rate for the entire evaluation period was 3%. Given these results, it can be concluded that this application has the potential to transform milk quality control and improve food safety.

Characteristics and Development Achievements of Modern Painting Art Based on the Perspective of Fine Arts in the Context of Digitization

Abstract Modernism has brought a new fine art vision to the art of painting, making it show aesthetic characteristics that distinguish it from classical painting. In this paper, based on the method of information theory, the indexes of order, complexity, and saliency of images are proposed to characterize the digital features of different aspects of the visual art of painting. The FisherScore selection mechanism is used to evaluate and screen the excellent features extracted. The final features are input into the Key Area Description Network (KADN) to complete the digital construction of the features of painting art through the KADN algorithm, and finally, based on the Wikiart dataset, we carry out the experiments of feature extraction and analysis of the images of fine art paintings, and summarize the achievements of the development of the art of painting through the results of the experiments. The results show that the improvement of KADN’s classification performance for the style subset is reflected in the whole. For example, the class Ukiyo-e only has a Top-1 classification error rate of about 12.2% under the key region description network, while the Top-1 classification error rate of the Realism class reaches more than 61.5%. Modern painting art can be analyzed effectively using the feature extraction method, which is a powerful tool for exploring the evolution of painting features.

Deep Multi-Layer Neural Network with Variable-Depth Output

In this study, a deep multi-layer neural network (DMLNN) with variable-depth output (VDO), called VDO-DMLNN, is proposed for classification. Unlike the traditional DMLNN, for which a user must define the network architecture in advance, VDO-DMLNN is produced from the top–down, layer by layer, until the classification error rate of VDO-DMLNN no longer decreases. The user thus does not need to define the depth of VDO-DMLNN in advance. The combination of the genetic algorithm (GA) and the self-organizing feature map (SOFM), called GA–SOFM, is proposed to automatically generate the weights and proper number of nodes for each layer in VDO-DMLNN. In addition, the output nodes can be at different levels in VDO-DMLNN rather than all being at the last layer, as in the traditional DMLNN. Thus, the average of computing time required for the recognition of an input sample in VDO-DMLNN is less than that in traditional DMLNN when they have the same classification error rate. Finally, VDO-DMLNN is compared with some state-of-the-art neural networks in the experiments.

Novel metabolic biomarkers for diagnosis of Alzheimer’s disease

AbstractBackgroundThe need for biomarkers for the etiological identification of dementia syndromes can not be overemphasized. Blood metabolites are attractive targets for identifying biomarkers with predictive values for AD diagnosis and progression. This study aims to identify novel AD blood metabolic biomarkers and develop a high‐sensitivity and specificity blood‐based test for AD.MethodIn the discovery phase, the serum metabolome of AD patients (n = 10) and age, sex‐matched healthy controls (n = 10) were analyzed with a mass spectrometry (MS)‐based approach. Supervised machine learning was used to identify the biomarker candidates and develop a computational model for AD diagnosis. Finally, the computational model and biomarker candidates were verified with a second AD cohort in the verification phase, where the serum metabolome of AD patients (n = 27) and age, sex‐matched healthy controls (n = 30) were analyzed.ResultIn this discovery study, 7075 metabolic features were quantified. Among them, the abundance of 750 unique features significantly differed between AD and Controls (t‐test p‐value &lt;0.05). Support vector machine (SVM) with radial basis function (RBF) kernel was used for features optimization and identified a panel of 14 metabolic features which predicted AD with 0.0% classification error rate (4‐fold cross‐validation). In the verification study, 10595 metabolic features were quantified. Among them, 908 unique features significantly differed between AD and Controls (t‐test with Permutation q‐value &lt;0.05). Then, the computational model and biomarker panel identified in the discovery phase were tested with the verification dataset and confirmed that they could predict AD with high accuracy (0.0% classification error rate, 4‐fold cross‐validation).ConclusionThis study comprehensively profiled the AD serum metabolome and identified a panel of metabolic biomarkers and a computational model for AD diagnosis, which will serve as a foundation for a high‐performance, blood‐based test for clinical AD screening and diagnosis.

AdaInject: Injection-Based Adaptive Gradient Descent Optimizers for Convolutional Neural Networks

The convolutional neural networks (CNNs) are generally trained using stochastic gradient descent (SGD) based optimization techniques. The existing SGD optimizers generally suffer with the overshooting of the minimum and oscillation near minimum. In this paper, we propose a new approach, hereafter referred as AdaInject, for the gradient descent optimizers by injecting the second order moment into the first order moment. Specifically, the short-term change in parameter is used as a weight to inject the second order moment in the update rule. The AdaInject optimizer controls the parameter update, avoids the overshooting of the minimum and reduces the oscillation near minimum. The proposed approach is generic in nature and can be integrated with any existing SGD optimizer. The effectiveness of the AdaInject optimizer is explained intuitively as well as through some toy examples. We also show the convergence property of the proposed injection based optimizer. Further, we depict the efficacy of the AdaInject approach through extensive experiments in conjunction with the state-of-the-art optimizers, namely AdamInject, diffGradInject, RadamInject, and AdaBeliefInject on four benchmark datasets. Different CNN models are used in the experiments. A highest improvement in the top-1 classification error rate of 16.54% is observed using diffGradInject optimizer with ResNeXt29 model over the CIFAR10 dataset. Overall, we observe very promising performance improvement of existing optimizers with the proposed AdaInject approach.

Thriving arid oasis urban agglomerations: Optimizing ecosystem services pattern under future climate change scenarios using dynamic Bayesian network

The effects of global climate change and human activities are anticipated to significantly impact ecosystem services (ESs), particularly in urban agglomerations of arid regions. This paper proposes a framework integrating the dynamic Bayesian network (DBN), system dynamics (SD) model, patch generation land use simulation (PLUS) model, and the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model for predicting land use change and optimizing ESs spatial patterns that is built upon the SSP-RCP scenarios from CMIP6. This framework is applied to the oasis urban agglomeration on the northern slope of the Tianshan Mountains in Xinjiang (UANSTM), China. The findings indicate that both the SD model and PLUS model can accurately forecast the distribution of future land use. The SD model shows a relative error of less than 2.32%, while the PLUS model demonstrates a Kappa coefficient of 0.89. The land use pattern displays obvious spatial heterogeneity under different climate scenarios. The expansion of cultivated land and construction land is the main form of land use change in UANSTM in the future. The DBN model proficiently simulates the interactive relationships between ESs and diverse factors. The classification error rates for net primary productivity (NPP), habitat quality (HQ), water yield (WY), and soil retention (SR) are 20.04%, 3.47%, 4.45%, and 13.42%, respectively. The prediction and diagnosis of DBN determine the optimal ESs development scenario and the optimal ESs region in the study area. It is found that the majority of ESs in UANSTM are predominantly influenced by natural factors with the exception of HQ. The socio-economic development plays a minor role in such urban agglomerations. This study offers significant insights that can contribute to the fields of ecological protection and land use planning in arid urban agglomerations worldwide.

Innovative Bacterial Colony Detection: Leveraging Multi-Feature Selection with the Improved Salp Swarm Algorithm.

In this paper, we introduce a new and advanced multi-feature selection method for bacterial classification that uses the salp swarm algorithm (SSA). We improve the SSA's performance by using opposition-based learning (OBL) and a local search algorithm (LSA). The proposed method has three main stages, which automate the categorization of bacteria based on their unique characteristics. The method uses a multi-feature selection approach augmented by an enhanced version of the SSA. The enhancements include using OBL to increase population diversity during the search process and LSA to address local optimization problems. The improved salp swarm algorithm (ISSA) is designed to optimize multi-feature selection by increasing the number of selected features and improving classification accuracy. We compare the ISSA's performance to that of several other algorithms on ten different test datasets. The results show that the ISSA outperforms the other algorithms in terms of classification accuracy on three datasets with 19 features, achieving an accuracy of 73.75%. Additionally, the ISSA excels at determining the optimal number of features and producing a better fit value, with a classification error rate of 0.249. Therefore, the ISSA method is expected to make a significant contribution to solving feature selection problems in bacterial analysis.

A random forest algorithm-based prediction model for moderate to severe acute postoperative pain after orthopedic surgery under general anesthesia

BackgroundPostoperative pain is one of the most common complications after surgery. In order to detect early and intervene in time for moderate to severe postoperative pain, it is necessary to identify risk factors and construct clinical prediction models. This study aimed to identify significant risk factors and establish a better-performing model to predict moderate to severe acute postoperative pain after orthopedic surgery under general anesthesia.MethodsPatients who underwent orthopedic surgery under general anesthesia were divided into patients with moderate to severe pain group (group P) and patients without moderate to severe pain group (group N) based on VAS scores. The features selected by Lasso regression were processed by the random forest and multivariate logistic regression models to predict pain outcomes. The classification performance of the two models was evaluated through the testing set. The area under the curves (AUC), the accuracy of the classifiers, and the classification error rate for both classifiers were calculated, the better-performing model was used to predict moderate to severe acute postoperative pain after orthopedic surgery under general anesthesia.ResultsA total of 327 patients were enrolled in this study (228 in the training set and 99 in the testing set). The incidence of moderate to severe postoperative pain was 41.3%. The random forest model revealed a classification error rate of 25.2% and an AUC of 0.810 in the testing set. The multivariate logistic regression model revealed a classification error rate of 31.3% and an AUC of 0.764 in the testing set. The random forest model was chosen for predicting clinical outcomes in this study. The risk factors with the greatest and second contribution were immobilization and duration of surgery, respectively.ConclusionsThe random forest model can be used to predict moderate to severe acute postoperative pain after orthopedic surgery under general anesthesia, which is of potential clinical application value.

Representation constraint‐based dual‐channel network for face antispoofing

AbstractAlthough multimodal face data have obvious advantages in describing live and spoofed features, single‐modality face antispoofing technologies are still widely used when it is difficult to obtain multimodal face images or inconvenient to integrate and deploy multimodal sensors. Since the live/spoofed representations in visible light facial images include considerable face identity information interference, existing deep learning‐based face antispoofing models achieve poor performance when only the visible light modality is used. To address the above problems, the authors design a dual‐channel network structure and a constrained representation learning method for face antispoofing. First, they design a dual‐channel attention mechanism‐based grouped convolutional neural network (CNN) to learn important deceptive cues in live and spoofed faces. Second, they design inner contrastive estimation‐based representation constraints for both live and spoofed samples to minimise the sample similarity loss to prevent the CNN from learning more facial appearance information. This increases the distance between live and spoofed faces and enhances the network's ability to identify deceptive cues. The evaluation results indicate that the framework we designed achieves an average classification error rate (ACER) of 2.37% on the visible light modality subset of the CASIA‐SURF dataset and an ACER of 2.4% on the CASIA‐SURF CeFA dataset, outperforming existing methods. The proposed method achieves low ACER scores in cross‐dataset testing, demonstrating its advantage in domain generalisation.

LRFID-Net: A Local-Region-Based Fake-Iris Detection Network for Fake Iris Images Synthesized by a Generative Adversarial Network

Iris recognition is a biometric method using the pattern of the iris seated between the pupil and the sclera for recognizing people. It is widely applied in various fields owing to its high accuracy in recognition and high security. A spoof detection method for discriminating a spoof attack is essential in biometric recognition systems that include iris recognition. However, previous studies have mainly investigated spoofing attack detection methods based on printed or photographed images, video replaying, artificial eyes, and patterned contact lenses fabricated using iris images from information leakage. On the other hand, there have only been a few studies on spoof attack detection using iris images generated through a generative adversarial network (GAN), which is a method that has drawn considerable research interest with the recent development of deep learning, and the enhancement of spoof detection accuracy by the methods proposed in previous research is limited. To address this problem, the possibility of an attack on a conventional iris recognition system with spoofed iris images generated using cycle-consistent adversarial networks (CycleGAN), which was the motivation of this study, was investigated. In addition, a local region-based fake-iris detection network (LRFID-Net) was developed. It provides a novel method for discriminating fake iris images by segmenting the iris image into three regions based on the iris region. Experimental results using two open databases, the Warsaw LiveDet-Iris-2017 and the Notre Dame Contact Lens Detection LiveDet-Iris-2017 datasets, showed that the average classification error rate of spoof detection by the proposed method was 0.03% for the Warsaw dataset and 0.11% for the Notre Dame Contact Lens Detection dataset. The results confirmed that the proposed method outperformed the state-of-the-art methods.

A Multimodal Transfer Learning Method for Classifying Images of Celestial Point Sources

A large fraction of celestial objects exhibit point shapes in CCD images, such as stars and QSOs, which contain less information due to their few pixels. Point source classification based solely on image data may lead to low accuracy. To address this challenge, this paper proposes a Multi-modal Transfer Learning-based classification method for celestial objects with point shape images. Considering that spectral data possess rich features and that there is a correlation between spectral data and image data, the proposed approach fully utilizes the knowledge gained from celestial spectral data and transfers it to the original image-based classification, enhancing the accuracy of classifying stars and QSOs. Initially, a one-dimensional residual network is employed to extract a 128-dimensional spectral feature vector from the original 3700-dimensional spectral data. This spectral feature vector captures important features of the celestial object. The Generative Adversarial Network is then utilized to generate a simulated spectral vector of 128 dimensions, which corresponds to the celestial object image. By generating simulated spectral vectors, data from two modals (spectral and image) for the same celestial object are available, enriching the input features of the model. In the upcoming multimodal classification model, we only require the images of celestial objects along with their corresponding simulated spectral data, and we no longer need real spectral data. With the assistance of spectral data, the proposed method alleviates the above disadvantages of the original image-based classification method. Remarkably, our method has improved the F1-score from 0.93 to 0.9777, while reducing the error rate in classification by 40%. These enhancements significantly increase the classification accuracy of stars and QSOs, providing strong support for the classification of celestial point sources.