Domain Classification Research Articles

Enhancing Food Image Recognition by Multi-Level Fusion and the Attention Mechanism

As a pivotal area of research in the field of computer vision, the technology for food identification has become indispensable across diverse domains including dietary nutrition monitoring, intelligent service provision in restaurants, and ensuring quality control within the food industry. However, recognizing food images falls within the domain of Fine-Grained Visual Classification (FGVC), which presents challenges such as inter-class similarity, intra-class variability, and the complexity of capturing intricate local features. Researchers have primarily focused on deep information in deep convolutional neural networks for fine-grained visual classification, often neglecting shallow and detailed information. Taking these factors into account, we propose a Multi-level Attention Feature Fusion Network (MAF-Net). Specifically, we use feature maps generated by the Convolutional Neural Networks (CNNs) backbone network at different stages as inputs. We apply a self-attention mechanism to identify local features on these feature maps and then stack them together. The feature vectors obtained through the attention mechanism are then integrated with the original input to enhance data augmentation. Simultaneously, to capture as many local features as possible, we encourage multi-scale features to concentrate on distinct local regions at each stage by maximizing the Kullback-Leibler Divergence (KL-divergence) between the different stages. Additionally, we present a novel approach called subclass center loss (SCloss) to implement label smoothing, minimize intra-class feature distribution differences, and enhance the model’s generalization capability. Experiments conducted on three food image datasets—CETH Food-101, Vireo Food-172, and UEC Food-100—demonstrated the superiority of the proposed model. The model achieved Top-1 accuracies of 90.22%, 89.86%, and 90.61% on CETH Food-101, Vireo Food-172, and UEC Food-100, respectively. Notably, our method not only outperformed other methods in terms of the Top-5 accuracy of Vireo Food-172 but also achieved the highest performance in the Top-1 accuracies of UEC Food-100.

Adapting performance metrics for ordinal classification to interval scale: length matters

In the field of supervised machine learning, accurate evaluation of classification models is a critical factor for assessing their performance and guiding model selection. This paper delves into the domain of ordinal classification and raises the question of adapting ordinal metrics to the interval scale. In scenarios where measurements are recorded at intervals, not only the order but also their length assume significance, and this promotes the adoption of novel performance metrics. Initially, we revisit two existing confusion matrix-based ordinal metrics and introduce a normalization technique to render them comparable and enhance their practical utility. We extend our focus to classification by intervals, proposing a robust framework for adapting ordinal metrics to the interval scale, and applying it to the aforementioned ordinal metrics. We address the challenge of unbounded rightmost intervals, a common issue in practical applications, from both theoretical and simulation perspectives, by providing a solution that enhances the applicability of the proposed metrics. To further explore practical implications, we conducted experiments on real-world datasets. The results reveal a promising trend in the use of interval-scale metrics to guide hyper-parameter tuning for improving model performance.

Deep Learning-Based Approach for Microscopic Algae Classification with Grad-CAM Interpretability

The natural occurrence of harmful algal blooms (HABs) adversely affects the quality of clean and fresh water. They pose increased risks to human health, aquatic ecosystems, and water bodies. Continuous monitoring and appropriate measures must be taken to combat HABs. Deep learning models that utilize computer vision play a vital role in identifying and classifying harmful algal blooms in aquatic environments and water storage facilities. Inspecting algal blooms using conventional methods, such as algae detection under microscopes, is difficult, expensive, and time-consuming. Deep learning algorithms have shown a notable and remarkable performance in the image classification domain and its applications, including microscopic algae species classification and detection. In this study, we propose a deep learning-based approach for classifying microscopic images of algae using computer vision. This approach employs a convolutional neural network (CNN) model integrated with two additional blocks—squeeze and dense blocks—to determine the presence of algae, followed by adding Grad-CAM to the proposed model to ensure interpretability and transparency. We performed several experiments on our custom dataset of microscopic algae images. Data augmentation techniques were employed to increase the number of images in the dataset, whereas pre-processing techniques were implemented to elevate the overall data quality. Our proposed model was trained on 3200 images consisting of four classes. We also compared our proposed model with the other transfer learning models, i.e., ResNet50 and Vgg16. Our proposed model outperformed the other two deep learning models. The proposed model demonstrated 96.7% accuracy, while Resnet50, EfficientNetB0, and VGG16 showed accuracy of 85.0%, 92.96%, and 93.5%, respectively. The results of this research demonstrate the potential of deep learning-based approaches for algae classification. This deep learning-based algorithm can be deployed in real-time applications to classify and identify algae to ensure the quality of water reservoirs. Computer-assisted solutions are advantageous for tracking freshwater algal blooms. Using deep learning-based models to identify and classify algae species from microscopic images is a novel application in the AI community.

High-intensive physical rehabilitation approach in children and adolescents with acquired brain injury during subacute phase (REHABILITY): a feasibility study protocol

IntroductionWhile principles of neuroplasticity and motor learning emphasise the potential of high dosage of physical rehabilitation in children and adolescents with acquired brain injury (ABI) during the subacute phase, we...

Evaluation of Different Few-Shot Learning Methods in the Plant Disease Classification Domain.

Early detection of plant diseases is crucial for agro-holdings, farmers, and smallholders. Various neural network architectures and training methods have been employed to identify optimal solutions for plant disease classification. However, research applying one-shot or few-shot learning approaches, based on similarity determination, to the plantdisease classification domain remains limited. This study evaluates different loss functions used in similarity learning, including Contrastive, Triplet, Quadruplet, SphereFace, CosFace, and ArcFace, alongside various backbone networks, such as MobileNet, EfficientNet, ConvNeXt, and ResNeXt. Custom datasets of real-life images, comprising over 4000 samples across 68 classes of plant diseases, pests, and their effects, were utilized. The experiments evaluate standard transfer learning approaches alongside similarity learning methods based on two classes of loss function. Results demonstrate the superiority of cosine-based methods over Siamese networks in embedding extraction for disease classification. Effective approaches for model organization and training are determined. Additionally, the impact of data normalization is tested, and the generalization ability of the models is assessed using a special dataset consisting of 400 images of difficult-to-identify plant disease cases.

A Target Domain-Specific Classifier Weight Partial Transfer Adversarial Network for Bearing Fault Diagnosis

In actual industry applications, the failure categories of practical equipment are usually a subset of laboratory conditions failure categories. Due to the strict constraints, partial transfer learning can address a more practical diagnostic scenario. In view of this, this paper proposes a target domain-specific classifier weight partial transfer adversarial network. Initially, the 1-D convolutional neural network is employed as the basic architecture. By training the domain discriminator and feature generator with an adversarial strategy, the recognition ability of the domain discriminant network and the feature extraction ability of the feature generation network can be enhanced. After that, a weighted learning strategy is introduced to guide the model to learn the cross-domain invariant feature. Also, a specific target domain classifier is utilized to redivide the target domain decision boundary to accurately classify the unlabeled target domain samples. Finally, five mainstream deep neural network methods are taken for comparison using the data from Western Reserve University and the motor-magnetic brake test designed by us. The results show that the proposed method reaches 90.18% and 96.53% classification accuracy on two datasets, respectively, which demonstrates superior performance compared with the state-of-the-art methods.

A Multi-Source Domain Adaptation Method for Bearing Fault Diagnosis with Dynamically Similarity Guidance on Incomplete Data

In actual industrial scenarios, collecting a complete dataset with all fault categories under the same conditions is challenging, leading to a loss in fault category knowledge in single-source domains. Deep learning domain adaptation methods face difficulties in multi-source scenarios due to insufficient labeled data and significant distribution differences, hindering domain-specific knowledge transfer and reducing fault diagnosis efficiency. To address these issues, the Dynamic Similarity-guided Multi-source Domain Adaptation Network (DS-MDAN) is proposed. This method leverages incomplete data from multiple-source domains to address distribution disparities in deep domain adaptation. It enhances diagnostic performance in the target domain by transferring knowledge across diverse domains. DS-MDAN uses convolution kernels of different scales to extract multi-scale feature information and achieves feature fusion through upsampling and operations like addition and concatenation. Adversarial training with domain and fault classifiers optimizes feature extraction for widely applicable representations. The similarity between source and target domain data is calculated based on features extracted by a shared-weight network, dynamically adjusting the contribution of different source domain data to minimize distribution differences. Finally, matched source and target domain samples are mapped to the same feature space for fault diagnosis. Experimental validation on various bearing fault datasets shows that DS-MDAN improves performance in multiple fault diagnosis tasks, increasing accuracy and demonstrating good generalization capabilities.

Video-Based Plastic Bag Grabbing Action Recognition: A New Video Dataset and a Comparative Study of Baseline Models.

Recognizing the action of plastic bag taking from CCTV video footage represents a highly specialized and niche challenge within the broader domain of action video classification. To address this challenge, our paper introduces a novel benchmark video dataset specifically curated for the task of identifying the action of grabbing a plastic bag. Additionally, we propose and evaluate three distinct baseline approaches. The first approach employs a combination of handcrafted feature extraction techniques and a sequential classification model to analyze motion and object-related features. The second approach leverages a multiple-frame convolutional neural network (CNN) to exploit temporal and spatial patterns in the video data. The third approach explores a 3D CNN-based deep learning model, which is capable of processing video data as volumetric inputs. To assess the performance of these methods, we conduct a comprehensive comparative study, demonstrating the strengths and limitations of each approach within this specialized domain.

A machine learning based variable selection algorithm for binary classification of perinatal mortality.

The identification of significant predictors with higher model performance is the key objective in classification domain. A machine learning-based variable selection technique termed as CARS-Logistic model is proposed by coupling competitive adaptive re-weighted sampling(CARS) and logistic regression for binary classification. Based on five assessment criteria, the proposed method is found to be more efficient than Forward selection logistic regression model. The CARS-Logistic model is executed to determine the significant factors of perinatal mortality in Pakistan. The identified hazards communicated social, cultural, financial, and health-related characteristics which contain key information about perinatal mortality in Pakistan for policymakers.

A milestone for the solution to the lattice sphere covering problem in dimension n = 6.

The complete classification of (primitive, generic) parallelohedra in a given dimension is a challenging computational task. Nearly 50 years have passed since the classification for the last dimension, n = 5, was completed. One application of such a classification is in solving the lattice sphere covering problem for the corresponding dimension. The paper by Dutour Sikirić & van Woerden [Acta Cryst. (2025), A81, https://doi.org/10.1107/S2053273324010143] marks a milestone in the classification effort for dimension n = 6. It provides a complete classification of all primitive iso-edge domains; here primitive parallelohedra are identified based on their facet vectors.

A comparative study of pre-trained models for image feature extraction in weather image classification using orange data mining

This paper presents a detailed comparative analysis of pre-trained models for feature extraction in the domain of weather image classification. Utilizing the orange data mining toolkit, we investigated the effectiveness of six prominent pre-trained models-InceptionV3, SqueezeNet, VGG-16, VGG-19, painter, and DeepLoc-in accurately classifying weather phenomena images. Among these models, InceptionV3, in conjunction with neural networks, emerged as the most effective, achieving a classification accuracy (CA) of 96.1%. Painter and SqueezeNet also showed strong performance, with accuracies of 95.1% and 86.7%, respectively, although they were surpassed by InceptionV3. VGG-16 and VGG-19 provided moderate accuracy, while DeepLoc underperformed significantly with a maximum accuracy of 56%. Neural networks consistently outperformed other classifiers across all models. This study highlights the critical importance of selecting appropriate pre-trained models to enhance the accuracy and reliability of weather image classification systems.

Intelligent Remote Sensing Scene Classification Model for On-Board Training of Resource-Constrained Devices

Remote Sensing Scene Classification (RSSC) is the distinctive classification of remote sensing images into numerous classes of scene classifications based on the image content. RSSC plays a significant role in several domains, like land mapping, agriculture, and the classification of disaster-prone regions. The Internet of Things (IoT) is a dynamic global network of devices, for example, vehicles, sensors, actuators, surveillance cameras, etc. These interconnected objects were distinctively recognizable and they could separately transfer and obtain valuable data through the network. However, satellite images were frequently degraded and blurred owing to aerosol dispersion under haze, fog, and other weather circumstances, decreasing the color fidelity and contrast of the image. To use effectual RSSC in real-time, widespread researchers concentrate on creating aerospace image processing systems, like airborne or spaceborne systems. Recently, with the quick improvement of deep learning (DL) and Machine learning (ML) techniques, the performance of RSSC has significantly developed owing to the hierarchical feature representation learning. Both technique has greater achievement in the domain of image scene classification. This study presents a Leveraging Tiny Convolutional Neural Networks with a Water Cycle Algorithm for Remote Sensing Scene Classification (LTCNN-WCRSSC) model. The LTCNN-WCRSSC technique is designed for efficient RSS classification in resource-constrained devices with on-board training capabilities. At first, the LTCNN-WCRSSC model applies image processing using a median filter (MF) to eliminate the noise. Next, the feature extraction process can be exploited by the ConvNeXt-Tiny method. For the RSSC model, the spatiotemporal attention bidirectional long short-term memory (STA-BiLSTM) technique is performed. Eventually, the water cycle algorithm (WCA)-based hyperparameter choice process can be performed to optimize the classification results of the STA-BiLSTM algorithm. The experimental evaluation of the LTCNN-WCRSSC technique takes place using a benchmark image dataset. The stimulated results indicated the superior performances of the LTCNN-WCRSSC model over other approaches.

Domain adaptive YOLO based on image style selection and synergistic domain classifier

Domain adaptive YOLO based on image style selection and synergistic domain classifier

DPAM-AI: a domain parser for AlphaFold models powered by artificial intelligence.

Due to the breakthrough in protein structure prediction by AlphaFold, the scientific community has access to 200 million predicted protein structures with near-atomic accuracy from the AlphaFold protein structure DataBase (AFDB), covering nearly the entire protein universe. Segmenting these models into domains and classifying them into an evolutionary hierarchy hold tremendous potential for unraveling essential insights into protein function. We introduce DPAM-AI, a Domain Parser for AlphaFold Models based on Artificial Intelligence. DPAM-AI utilizes a convolutional neural network trained with previously classified domains in the Evolutionary Classification Of protein Domains (ECOD) database. DPAM-AI integrates inter-residue distances, predicted aligned errors, and sequence and structural alignments to previously classified domains detected via sequence (HHsuite) and structural (Dali) similarity searches. DPAM-AI has demonstrated its power through rigorous tests, excelling in several benchmark sets compared to its predecessor, DPAM, and other recently published domain parsers, Merizo and Chainsaw. We applied DPAM-AI to representative AFDB models for proteins classified in Pfam. We obtained representative 3D structures for 18487 (89%) of the 20795 Pfam families. The remaining families either (i) belong to viral proteins that were excluded from AFDB or (ii) do not adopt globular 3D structures. Our structure-aware domain delineation uncovered a considerable fraction (15%) of Pfam domains containing multiple structural and evolutionary units and refined the boundaries for over half. Pfam and corresponding DPAM-AI domains are at http://prodata.swmed.edu/DPAM-pfam/. Our code is deposited at https://github.com/Jsauce5p/DPAM/tree/dpam_ai, and updates will be released through https://github.com/CongLabCode/DPAM.

SR-DAYOLOv8: cross-domain adaptive object detection based on super-resolution domain classifier

SR-DAYOLOv8: cross-domain adaptive object detection based on super-resolution domain classifier

MDRN: Multi‐domain representation network for unsupervised domain generalization

AbstractIn deep neural networks, performance can degrade when test data distributions differ from training data. Unsupervised Domain Generalization (UDG) aims to improve generalization across unseen domains by leveraging multiple source domains without supervision. Traditional methods focus on extracting domain‐invariant features, potentially at the expense of feature space integrity and generalization potential. We presents a Multi‐Domain Representation Network (MDRN) for unsupervised multi‐domain learning. MDRN innovates by disentangling and preserving both domain‐invariant and domain‐specific features through an unsupervised cross‐domain reconstruction task. It employs content encoders for domain‐invariant features and multi‐domain style encoders for domain‐specific characteristics. By merging these features based on domain similarity, MDRN constructs a comprehensive feature space that enhances image reconstruction across domains. Additionally, MDRN integrates domain‐specific classifiers, which learn domain classification and provide weighted fusion of domain‐specific features. This design facilitates effective inter‐domain distance measurement and feature integration. Experiments on PACS and DomainNet show MDRN's superior performance over existing state‐of‐the‐art UDG approaches, highlighting its effectiveness in handling distribution shifts between source and target domains.

SKIN RASH CLASSIFICATION SYSTEM USING MODIFIED DENSENET201 THROUGH RANDOM SEARCH FOR HYPERPARAMETER TUNING

Skin rashes caused by various diseases, such as monkeypox, cowpox, chickenpox, measles, and HFMD, often present similar symptoms, making accurate diagnosis challenging. This study aims to improve the classification of skin diseases through the application of a modified DenseNet-201 architecture combined with hyperparameter optimization using Random Search. The base DenseNet-201 model, with pre-trained weights, was first tested, achieving an accuracy of 63%, with the highest performance in the Healthy and HFMD classes. The proposed modified model, optimized using Random Search, improved overall accuracy to 80%, with enhanced precision, recall, and F1-score across most classes. The model’s performance was particularly notable in the HFMD and normal skin classes, although further improvements are needed for challenging classes like Cowpox and Measles. The findings highlight the potential of Random Search for hyperparameter tuning to enhance the performance of deep convolutional neural networks in the medical image classification domain, offering a promising tool for efficient and accurate skin disease detection.

Uncovering the Top Nonadvertising Weight Loss Websites on Google: A Data-Mining Approach.

Online weight loss information is commonly sought by internet users, and it may impact their health decisions and behaviors. Previous studies examined a limited number of Google search queries and relied on manual approaches to retrieve online weight loss websites. This study aimed to identify and describe the characteristics of the top weight loss websites on Google. This study gathered 432 Google search queries collected from Google autocomplete suggestions, "People Also Ask" featured questions, and Google Trends data. A data-mining software tool was developed to retrieve the search results automatically, setting English and the United States as the default criteria for language and location, respectively. Domain classification and evaluation technologies were used to categorize the websites according to their content and determine their risk of cyberattack. In addition, the top 5 most frequent websites in nonadvertising (ie, nonsponsored) search results were inspected for quality. The results revealed that the top 5 nonadvertising websites were healthline.com, webmd.com, verywellfit.com, mayoclinic.org, and womenshealthmag.com. All provided accuracy statements and author credentials. The domain categorization taxonomy yielded a total of 101 unique categories. After grouping the websites that appeared less than 5 times, the most frequent categories involved "Health" (104/623, 16.69%), "Personal Pages and Blogs" (91/623, 14.61%), "Nutrition and Diet" (48/623, 7.7%), and "Exercise" (34/623, 5.46%). The risk of being a victim of a cyberattack was low. The findings suggested that while quality information is accessible, users may still encounter less reliable content among various online resources. Therefore, better tools and methods are needed to guide users toward trustworthy weight loss information.

Innovative CNN approach for reliable chicken meat classification in the poultry industry

In response to the burgeoning need for advanced object recognition and classification, this research embarks on a journey harnessing the formidable capabilities of Convolutional Neural Networks (CNNs). The central aim of this study revolves around the precise identification and categorization of objects, with a specific focus on the critical task of distinguishing between fresh and spoiled chicken meat. This study's overarching objective is to craft a robust CNN-based classification model that excels in discriminating between objects. In the context of our research, we set out to create a model adept at distinguishing between fresh and rotten chicken meat. This endeavor holds immense potential in augmenting food safety and elevating quality control standards within the poultry industry. Our research methodology entails meticulous data collection, which includes acquiring high-resolution images of chicken meat. This meticulously curated dataset serves as the bedrock for both training and testing our CNN model. To optimize the model, we employ the 'adam' optimizer, while critical performance metrics, such as accuracy, precision, recall, and the F1-score, are methodically computed to evaluate the model's effectiveness. Our experimental findings unveil the remarkable success of our CNN model, with consistent accuracy, precision, and recall metrics all reaching an impressive pinnacle of 94%. These metrics underscore the model's excellence in the realm of object classification, with a particular emphasis on its proficiency in distinguishing between fresh and rotten chicken meat. In summation, our research concludes that the CNN model has exhibited exceptional prowess in the domains of object recognition and classification. The model's high accuracy signifies its precision in furnishing accurate predictions, while its elevated precision and recall values accentuate its effectiveness in differentiating between object classes. Consequently, the CNN model stands as a robust foundation for future strides in object classification technology. As we peer into the horizon of future research, myriad opportunities beckon. Our CNN model's applicability extends beyond chicken meat classification, inviting exploration across diverse domains. Furthermore, the model's refinement and adaptation for specific challenges represent an exciting avenue for future work, promising heightened performance across a broader spectrum of object recognition tasks.

Gradient Harmonization in Unsupervised Domain Adaptation.

Unsupervised domain adaptation (UDA) intends to transfer knowledge from a labeled source domain to an unlabeled target domain. Many current methods focus on learning feature representations that are both discriminative for classification and invariant across domains by simultaneously optimizing domain alignment and classification tasks. However, these methods often overlook a crucial challenge: the inherent conflict between these two tasks during gradient-based optimization. In this paper, we delve into this issue and introduce two effective solutions known as Gradient Harmonization, including GH and GH++, to mitigate the conflict between domain alignment and classification tasks. GH operates by altering the gradient angle between different tasks from an obtuse angle to an acute angle, thus resolving the conflict and trade-offing the two tasks in a coordinated manner. Yet, this would cause both tasks to deviate from their original optimization directions. We thus further propose an improved version, GH++, which adjusts the gradient angle between tasks from an obtuse angle to a vertical angle. This not only eliminates the conflict but also minimizes deviation from the original gradient directions. Finally, for optimization convenience and efficiency, we evolve the gradient harmonization strategies into a dynamically weighted loss function using an integral operator on the harmonized gradient. Notably, GH/GH++ are orthogonal to UDA and can be seamlessly integrated into most existing UDA models. Theoretical insights and experimental analyses demonstrate that the proposed approaches not only enhance popular UDA baselines but also improve recent state-of-the-art models.