Top-1 Accuracy Research Articles

Attention-enhanced multimodal feature fusion network for clothes-changing person re-identification

Clothes-Changing Person Re-Identification is a challenging problem in computer vision, primarily due to the appearance variations caused by clothing changes across different camera views. This poses significant challenges to traditional person re-identification techniques that rely on clothing features. These challenges include the inconsistency of clothing and the difficulty in learning reliable clothing-irrelevant local features. To address this issue, we propose a novel network architecture called the Attention-Enhanced Multimodal Feature Fusion Network (AE-Net). AE-Net effectively mitigates the impact of clothing changes on recognition accuracy by integrating RGB global features, grayscale image features, and clothing-irrelevant features obtained through semantic segmentation. Specifically, global features capture the overall appearance of the person; grayscale image features help eliminate the interference of color in recognition; and clothing-irrelevant features derived from semantic segmentation enforce the model to learn features independent of the person’s clothing. Additionally, we introduce a multi-scale fusion attention mechanism that further enhances the model’s ability to capture both detailed and global structures, thereby improving recognition accuracy and robustness. Extensive experimental results demonstrate that AE-Net outperforms several state-of-the-art methods on the PRCC and LTCC datasets, particularly in scenarios with significant clothing changes. On the PRCC and LTCC datasets, AE-Net achieves Top-1 accuracy rates of 60.4% and 42.9%, respectively.

Code Recommendation for Schema Evolution of Mimic Storage Systems

Schema evolution of mimic storage systems is a time-consuming and error-prone task due to the redundant development of heterogeneous executors. The ORM-based proxy requires an entire class to represent the structure of a data table. There lacks domain-specific code recommendation techniques to boost storage development. To address this issue, we design a novel type of code context, i.e. schema context, that combines features of code text, syntax and structure. Regarding the requirements of class-level granularity, we focus on behavior and attribute in code syntax, and use element position and structural metrics to mine the hidden relationships. Based on schema context and an existing inference mode, we propose SchemaRec to recommend ORM-related class for the database executors once one of them has been changed. We conduct experiments with 110 open-source projects, and the results show that SchemaRec obtains more accurate results than Lucene, DeepCS, QobCS and SEA in terms of Top-1, Top-10 and MRR accuracy due to the better ability of context representation. We also find that code syntax is the most important information because it involves behavior and attribute information of ORM-related classes.

Identifying Key Factors for Predicting the Age at Peak Height Velocity in Preadolescent Team Sports Athletes Using Explainable Machine Learning

Maturation is a key factor in sports participation and often determines the young athletes’ characterization as a talent. However, there is no evidence of practical models for understanding the factors that discriminate children according to maturity. Hence, this study aims to deepen the understanding of the factors that affect maturity in 11-year-old Team Sports Athletes by utilizing explainable artificial intelligence (XAI) models. We utilized three established machine learning (ML) classifiers and applied the Sequential Forward Feature Selection (SFFS) algorithm to each. In this binary classification task, the logistic regression (LR) classifier achieved a top accuracy of 96.67% using the seven most informative factors (Sitting Height, Father’s Height, Body Fat, Weight, Height, Left and Right-Hand Grip Strength). The SHapley Additive exPlanations (SHAP) model was instrumental in identifying the contribution of each factor, offering key insights into variable importance. Independent sample t-tests on these selected factors confirmed their significance in distinguishing between the two classes. By providing detailed and personalized insights into child development, this integration has the potential to enhance the effectiveness of maturation prediction significantly. These advancements could lead to a transformative approach in young athletes’ pediatric growth analysis, fostering better sports performance and developmental outcomes for children.

ConvAtt Network: A Low Parameter Approach For Sign Language Recognition

Despite recent advances in Large Language Models in text processing, Sign Language Recognition (SLR) remains an unresolved task. This is, in part, due to limitations in the available data. In this paper, we investigate combining 1D convolutions with transformer layers to capture local features and global interactions in a low-parameter SLR model. We experimented using multiple data augmentation and regularization techniques to categorize signs of the French Belgian Sign Language. We achieved a top-1 accuracy of 42.7% and a top-10 accuracy of 81.9% in 600 different signs. This model is competitive with the current state of the art while using a significantly lower number of parameters.

Method for Enhancing AI Accuracy in Pressure Injury Detection Using Real and Synthetic Datasets

Pressure injuries pose significant health risks, especially for the elderly, immobile individuals, and those with sensory impairments. These injuries can rapidly become chronic, making initial diagnosis important. Due to the difficulty of transporting patients from local health facilities to higher-level general hospitals for treatment, it is essential to utilize telemedicine tools, such as chatbots, to ensure rapid initial diagnosis. Recent advances in artificial intelligence have demonstrated potential for medical imaging and disease classification. Ongoing research in the field of dermatological diseases focuses on disease classification. However, the assessment accuracy of artificial intelligence is often limited by unequal class distributions and insufficient dataset quantities. In this study, we aim to enhance the accuracy of artificial intelligence models by generating synthetic datasets. Specifically, we focused on training models for Pressure Injury assessment using both real and synthetic datasets. We used PI data at a domestic medical university. As part of our supplementary research, we established a chatbot system to facilitate the assessment of pressure injuries. Using both constructed and synthetic data, we achieved a top-1 accuracy of 92.03%. The experimental results demonstrate that combining real and synthetic data significantly improves model accuracy. These findings suggest that synthetic datasets can be effectively utilized to address the limitations of small-scale datasets in medical applications. Future research should explore the use of diverse synthetic data generation methods and validate model performance on a variety of datasets to enhance the generalization and robustness of AI models for Pressure Injury assessment.

CMSSP: A Contrastive Mass Spectra-Structure Pretraining Model for Metabolite Identification.

A pivotal challenge in metabolite research is the structural annotation of metabolites from tandem mass spectrometry (MS/MS) data. The integration of artificial intelligence (AI) has revolutionized the interpretation of MS data, facilitating the identification of elusive metabolites within the metabolomics landscape. Innovative methodologies are primarily focusing on transforming MS/MS spectra or molecular structures into a unified modality to enable similarity-based comparison and interpretation. In this work, we present CMSSP, a novel Contrastive Mass Spectra-Structure Pretraining framework designed for metabolite annotation. The primary objective of CMSSP is to establish a representation space that facilitates a direct comparison between MS/MS spectra and molecular structures, transcending the limitations of distinct modalities. The evaluation on two benchmark test sets demonstrates the efficacy of the approach. CMSSP achieved a remarkable enhancement in annotation accuracy, outperforming the state-of-the-art methods by a significant margin. Specifically, it improved the top-1 accuracy by 30% on the CASMI 2017 data set and realized a 16% increase in top-10 accuracy on an independent test set. Moreover, the model displayed superior identification accuracy across all seven chemical categories, showcasing its robustness and versatility. Finally, the MS/MS data of 30 metabolites from Glycyrrhiza glabra were analyzed, achieving top-1 and top-3 accuracies of 86.7 and 100%, respectively. The CMSSP model serves as a potent tool for the dissection and interpretation of intricate MS/MS data, propelling the field toward more accurate and efficient metabolite annotation. This not only augments the analytical capabilities of metabolomics but also paves the way for future discoveries in understanding of complex biological systems.

Nutritional composition analysis in food images: an innovative Swin Transformer approach.

Accurate recognition of nutritional components in food is crucial for dietary management and health monitoring. Current methods often rely on traditional chemical analysis techniques, which are time-consuming, require destructive sampling, and are not suitable for large-scale or real-time applications. Therefore, there is a pressing need for efficient, non-destructive, and accurate methods to identify and quantify nutrients in food. In this study, we propose a novel deep learning model that integrates EfficientNet, Swin Transformer, and Feature Pyramid Network (FPN) to enhance the accuracy and efficiency of food nutrient recognition. Our model combines the strengths of EfficientNet for feature extraction, Swin Transformer for capturing long-range dependencies, and FPN for multi-scale feature fusion. Experimental results demonstrate that our model significantly outperforms existing methods. On the Nutrition5k dataset, it achieves a Top-1 accuracy of 79.50% and a Mean Absolute Percentage Error (MAPE) for calorie prediction of 14.72%. On the ChinaMartFood109 dataset, the model achieves a Top-1 accuracy of 80.25% and a calorie MAPE of 15.21%. These results highlight the model's robustness and adaptability across diverse food images, providing a reliable and efficient tool for rapid, non-destructive nutrient detection. This advancement supports better dietary management and enhances the understanding of food nutrition, potentially leading to more effective health monitoring applications.

Techniques for Detecting the Start and End Points of Sign Language Utterances to Enhance Recognition Performance in Mobile Environments

Recent AI-based technologies in mobile environments have enabled sign language recognition, allowing deaf individuals to communicate effectively with hearing individuals. However, varying computational performance across different mobile devices can result in differences in the number of image frames extracted in real time during sign language utterances. The number of extracted frames is a critical factor influencing the accuracy of sign language recognition models. If the number of extracted frames is too small, the performance of the sign language recognition model may decline. Additionally, detecting the start and end points of sign language utterances is crucial for improving recognition accuracy, as the period before the start point and after the end point often involves no action being performed. These parts do not capture the unique characteristics of each sign language. Therefore, this paper proposes a technique to dynamically adjust the sampling rate based on the number of frames extracted in real time during sign language utterances in mobile environments, with the aim of accurately detecting the start and end points of the sign language. Experiments were conducted to compare the proposed technique with the fixed sampling rate method and with the no-sampling method as a baseline. Our findings show that the proposed dynamic sampling rate adjustment method improves performance by up to 83.64% in top-5 accuracy and by up to 66.54% in top-1 accuracy compared to the fixed sampling rate method. The performance evaluation results underscore the effectiveness of our dynamic sampling rate adjustment approach in enhancing the accuracy and robustness of sign language recognition systems across different operational conditions.

GestaltMatcher Database - A global reference for facial phenotypic variability in rare human diseases.

The most important factor that complicates the work of dysmorphologists is the significant phenotypic variability of the human face. Next-Generation Phenotyping (NGP) tools that assist clinicians with recognizing characteristic syndromic patterns are particularly challenged when confronted with patients from populations different from their training data. To that end, we systematically analyzed the impact of genetic ancestry on facial dysmorphism. For that purpose, we established the GestaltMatcher Database (GMDB) as a reference dataset for medical images of patients with rare genetic disorders from around the world. We collected 10,980 frontal facial images - more than a quarter previously unpublished - from 8,346 patients, representing 581 rare disorders. Although the predominant ancestry is still European (67%), data from underrepresented populations have been increased considerably via global collaborations (19% Asian and 7% African). This includes previously unpublished reports for more than 40% of the African patients. The NGP analysis on this diverse dataset revealed characteristic performance differences depending on the composition of training and test sets corresponding to genetic relatedness. For clinical use of NGP, incorporating non-European patients resulted in a profound enhancement of GestaltMatcher performance. The top-5 accuracy rate increased by +11.29%. Importantly, this improvement in delineating the correct disorder from a facial portrait was achieved without decreasing the performance on European patients. By design, GMDB complies with the FAIR principles by rendering the curated medical data findable, accessible, interoperable, and reusable. This means GMDB can also serve as data for training and benchmarking. In summary, our study on facial dysmorphism on a global sample revealed a considerable cross ancestral phenotypic variability confounding NGP that should be counteracted by international efforts for increasing data diversity. GMDB will serve as a vital reference database for clinicians and a transparent training set for advancing NGP technology.

Experimental assessment of the performance of artificial intelligence in solving multiple-choice board exams in cardiology.

The aim of the present study was to evaluate the performance of various artificial intelligence (AI)-powered chatbots (commercially available in Switzerland up to June 2023) in solving a theoretical cardiology board exam and to compare their accuracy with that of human cardiology fellows. For the study, a set of 88 multiple-choice cardiology exam questions was used. The participating cardiology fellows and selected chatbots were presented with these questions. The evaluation metrics included Top-1 and Top-2 accuracy, assessing the ability of chatbots and fellows to select the correct answer. Among the cardiology fellows, all 36 participants successfully passed the exam with a median accuracy of 98% (IQR 91-99%, range from 78% to 100%). However, the performance of the chatbots varied. Only one chatbot, Jasper quality, achieved the minimum pass rate of 73% correct answers. Most chatbots demonstrated a median Top-1 accuracy of 47% (IQR 44-53%, range from 42% to 73%), while Top-2 accuracy provided a modest improvement, resulting in a median accuracy of 67% (IQR 65-72%, range from 61% to 82%). Even with this advantage, only two chatbots, Jasper quality and ChatGPT plus 4.0, would have passed the exam. Similar results were observed when picture-based questions were excluded from the dataset. Overall, the study suggests that most current language-based chatbots have limitations in accurately solving theoretical medical board exams. In general, currently widely available chatbots fell short of achieving a passing score in a theoretical cardiology board exam. Nevertheless, a few showed promising results. Further improvements in artificial intelligence language models may lead to better performance in medical knowledge applications in the future.

NAS-BNN: Neural Architecture Search for Binary Neural Networks

Binary Neural Networks (BNNs) have gained extensive attention for their superior inferencing efficiency and compression ratio compared to traditional full-precision networks. However, due to the unique characteristics of BNNs, designing a powerful binary architecture is challenging and often requires significant manpower. A promising solution is to utilize Neural Architecture Search (NAS) to assist in designing BNNs, but current NAS methods for BNNs are relatively straightforward and leave a performance gap between the searched models and manually designed ones. To address this gap, we propose a novel neural architecture search scheme for binary neural networks, named NAS-BNN. We first carefully design a search space based on the unique characteristics of BNNs. Then, we present three training strategies, which significantly enhance the training of supernet and boost the performance of all subnets. Our discovered binary model family outperforms previous BNNs for a wide range of operations (OPs) from 20M to 200M. For instance, we achieve 68.20% top-1 accuracy on ImageNet with only 57M OPs. In addition, we validate the transferability of these searched BNNs on the object detection task, and our binary detectors with the searched BNNs achieve a novel state-of-the-art result, e.g., 31.6% mAP with 370M OPs, on MS COCO dataset. The source code and models will be released at https://github.com/VDIGPKU/NAS-BNN.

Interactive Surgical Training in Neuroendoscopy: Real-Time Anatomical Feature Localization Using Natural Language Expressions.

This study addresses challenges in surgical education, particularly in neuroendoscopy, where the demand for optimized workflow conflicts with the need for trainees' active participation in surgeries. To overcome these challenges, we propose a framework that accurately identifies anatomical structures within images guided by language descriptions, facilitating authentic and interactive learning experiences in neuroendoscopy. Utilizing the encoder-decoder architecture of a conventional transformer, our framework processes multimodal inputs (images and language descriptions) to identify and localize features in neuroendoscopic images. We curate a dataset from recorded endoscopic third ventriculostomy (ETV) procedures for training and evaluation. Utilizing evaluation metrics, including "R@n," "IoU= θ," "mIoU," and top-1 accuracy, we systematically benchmark our framework against state-of-the-art methodologies. The framework demonstrates excellent generalization, surpassing the compared methods with 93.67 % accuracy and 76.08 % mIoU on unseen data. It also exhibits better computational speed compared with other methods. Qualitative results affirms the framework's effectiveness in precise localization of referred anatomical features within neuroendoscopic images. The framework's adeptness at localizing anatomical features using language descriptions positions it as a valuable tool for integration into future interactive clinical learning systems, enhancing surgical training in neuroendoscopy. The exemplary performance reinforces the framework's potential in enhancing surgical education, leading to improved skills and outcomes for trainees in neuroendoscopy.

HoopTransformer: Advancing NBA Offensive Play Recognition with Self-Supervised Learning from Player Trajectories.

Understanding and recognizing basketball offensive set plays, which involve intricate interactions between players, have always been regarded as challenging tasks for untrained humans, not to mention machines. In this study, our objective is to propose an artificial intelligence model that can automatically recognize offensive plays using a novel self-supervised learning approach. The dataset was collected by SportVU from 632 games during the 2015-2016 season of the National Basketball Association (NBA), with a total of 90,524 possessions. A multi-agent motion prediction pretraining model was built on the basis of axial-attention transformer and trained with different masking strategies: motion prediction (MP), motion reconstruction (MR), and MP+MR joint strategy. A downstream play-level classification task and similarity search were used to evaluate the models' performance. The results showed that the MP+MR joint masking strategy maximized the ability of the model compared with individual masking strategies. For the classification task, the joint strategy achieved a top-1 accuracy of 81.5% and top-3 accuracy of 97.5%. In the similarity search evaluation, the joint strategy attained a top-5 accuracy of 76% and top-10 accuracy of 59%. Additionally, with the same MP+MR joint masking strategy, our HoopTransformer model outperformed the two baseline models in the classification task and similarity search. This study presents a self-supervised learning model and demonstrates the effectiveness and potential of the model in accurately comprehending and capturing player movements and complex interactions during offensive plays.

Visualizing and Understanding Patch Interactions in Vision Transformer.

Vision transformer (ViT) has become a leading tool in various computer vision tasks, owing to its unique self-attention mechanism that learns visual representations explicitly through cross-patch information interactions. Despite having good success, the literature seldom explores the explainability of ViT, and there is no clear picture of how the attention mechanism with respect to the correlation across comprehensive patches will impact the performance and what is the further potential. In this work, we propose a novel explainable visualization approach to analyze and interpret the crucial attention interactions among patches for ViT. Specifically, we first introduce a quantification indicator to measure the impact of patch interaction and verify such quantification on attention window design and indiscriminative patches removal. Then, we exploit the effective responsive field of each patch in ViT and devise a window-free transformer (WinfT) architecture accordingly. Extensive experiments on ImageNet demonstrate that the exquisitely designed quantitative method is shown able to facilitate ViT model learning, leading the top-1 accuracy by 4.28% at most. More remarkably, the results on downstream fine-grained recognition tasks further validate the generalization of our proposal.

Visual-guided hierarchical iterative fusion for multi-modal video action recognition

Vision-Language models(VLMs) have shown promising improvements on various visual tasks. Most existing VLMs employ two separate transformer-based encoders, each dedicated to modeling visual and language features independently. Because the visual features and language features are unaligned in the feature space, it is challenging for the multi-modal encoder to learn vision-language interactions. In this paper, we propose a V,isual-guided Hierarchical Iterative Fusion (VgHIF) method for VLMs in video action recognition, which acquires more discriminative vision and language representation. VgHIF leverages visual features from different levels in visual encoder to interact with language representation. The interaction is processed by the attention mechanism to calculate the correlation between visual features and language representation. VgHIF learns grounded video-text representation and supports many different pre-trained VLMs in a flexible and efficient manner with a tiny computational cost. We conducted experiments on the Kinetics-400 Mini Kinetics 200 HMDB51, and UCF101 using VLMs: CLIP, X-CLIP, and ViFi-CLIP. The experiments were conducted under full supervision and few shot settings, and compared with the baseline multi-modal model without VgHIF, the Top-1 accuracy of the proposed method has been improved to varying degrees, and several groups of results have achieved comparable results with state-of-the-art performance, which strongly verified the effectiveness of the proposed method.

ECTFormer: An efficient Conv-Transformer model design for image recognition

Since the success of Vision Transformers (ViTs), there has been growing interest in combining ConvNets and Transformers in the computer vision community. While the hybrid models have demonstrated state-of-the-art performance, many of these models are too large and complex to be applied to edge devices for real-world applications. To address this challenge, we propose an efficient hybrid network called ECTFormer that leverages the strengths of ConvNets and Transformers while considering both model performance and inference speed. Specifically, our approach involves: (1) optimizing the combination of convolution kernels by dynamically adjusting kernel sizes based on the scale of feature tensors; (2) revisiting existing overlapping patchify to not only reduce the model size but also propagate fine-grained patches for the performance enhancement; and (3) introducing an efficient single-head self-attention mechanism, rather than multi-head self-attention in the base Transformer, to minimize the increase in model size and boost inference speed, overcoming bottlenecks of ViTs. In experimental results on ImageNet-1K, ECTFormer not only demonstrates comparable or higher top-1 accuracy but also faster inference speed on both GPUs and edge devices compared to other efficient networks.

Accelerating one-shot neural architecture search via constructing a sparse search space

Neural Architecture Search (NAS) has garnered significant attention for its ability to automatically design high-quality deep neural networks (DNNs) tailored to various hardware platforms. The major challenge for NAS is the time-consuming network estimation process required to select optimal networks from a large pool of candidates. Rather than training each candidate from scratch, recent one-shot NAS methods accelerate the estimation process by only training a supernet and sampling sub-networks from it, inheriting partial network architectures and weights. Despite significant acceleration, the supernet training with a large search space (i.e., the number of candidate sub-networks) still requires thousands of GPU hours to support high-quality sub-network sampling. In this work, we propose SparseNAS, an approach for one-shot NAS acceleration by reducing the redundancy of the search space. We observe that many sub-networks in the space are underperforming, with significant performance disparity to high-performance sub-networks. Crucially, this disparity can be observed early in the beginning of the supernet training. Therefore, we train an early predictor to learn this disparity and filter out high-quality networks in advance. Then, the supernet training will be conducted in this space sub-space. Compared to the state-of-the-art one-shot NAS, our SparseNAS reports a 3.1× training speedup with comparable network performance on the ImageNet dataset. Compared to the state-of-the-art acceleration method, SparseNAS reports a maximum of 1.5% higher Top-1 accuracy and 28% training cost reduction with a 7× bigger search space. Extensive experiment results demonstrated that SparseNAS achieves better trade-offs between efficiency and performance than state-of-the-art one-shot NAS.

DINO-Mix enhancing visual place recognition with foundational vision model and feature mixing

Using visual place recognition (VPR) technology to ascertain the geographical location of publicly available images is a pressing issue. Although most current VPR methods achieve favorable results under ideal conditions, their performance in complex environments, characterized by lighting variations, seasonal changes, and occlusions, is generally unsatisfactory. Therefore, obtaining efficient and robust image feature descriptors in complex environments is a pressing issue. In this study, we utilized the DINOv2 model as the backbone for trimming and fine-tuning to extract robust image features and employed a feature mix module to aggregate image features, resulting in globally robust and generalizable descriptors that enable high-precision VPR. We experimentally demonstrated that the proposed DINO-Mix outperforms the current state-of-the-art (SOTA) methods. Using test sets having lighting variations, seasonal changes, and occlusions such as Tokyo24/7, Nordland, and SF-XL-Testv1, our proposed architecture achieved Top-1 accuracy rates of 91.75%, 80.18%, and 82%, respectively, and exhibited an average accuracy improvement of 5.14%. In addition, we compared it with other SOTA methods using representative image retrieval case studies, and our architecture outperformed its competitors in terms of VPR performance. Furthermore, we visualized the attention maps of DINO-Mix and other methods to provide a more intuitive understanding of their respective strengths. These visualizations serve as compelling evidence of the superiority of the DINO-Mix framework in this domain.

Sports action recognition algorithm based on multi-modal data recognition

The recognition of sports action is an important research subject, which is conducive to the improvement of athletes’ own level. To improve the accuracy of multi-modal data action recognition, based on the Transformer module, this study introduces a multi-head attention mechanism, fuses multi-modal data, and constructs a multi-stream structured object relationship inference network. Based on PointNet++ network and combining five different data fusion frameworks, a motion recognition model that integrates RGB data and 3D skeleton point cloud is constructed. The results showed that the Top-1 accuracy of multi-stream structured object relationship inference network was 42.5% and 42.7%, respectively, which was better than other algorithms. The accuracy of the multi-modal fusion model was improved by 15.6% and 5.1% compared with the single mode, and by 5.4% and 2.6% compared with the dual mode, which showed its superiority in the action recognition task. This showed that the fusion of multi-modal data can provide more abundant information, so as to improve the accuracy of action recognition. The accuracy of the action recognition model combining RGB data and 3D skeleton point cloud was 84.3%, 87.5%, 90.2%, 90.6% and 91.2% after the combination of different strategies, which effectively compensated for the problem of missing information in 3D skeleton point cloud and significantly improved the accuracy of action recognition. With a small amount of data, the Top-1 accuracy of the multi-stream structured object relationship inference network in this study was superior to other algorithms, showing its advantages in dealing with complex action recognition tasks. In addition, the action recognition model that fuses RGB data and 3D skeleton point cloud also achieved higher accuracy, which is better than other algorithms. This study can meet the needs of motion recognition in different scenarios and has certain reference value.

StAlK: Structural Alignment based Self Knowledge distillation for Medical Image Classification

In the realm of medical image analysis, where challenges like high class imbalance, inter-class similarity, and intra-class variance are prevalent, knowledge distillation has emerged as a powerful mechanism for model compression and regularization. Existing methodologies, including label smoothening, contrastive learning, and relational knowledge transfer, aim to address these challenges but exhibit limitations in effectively managing either class imbalance or intricate inter and intra-class relations within input samples. In response, this paper introduces StAlK (Structural Alignment based Self Knowledge distillation) for Medical Image Classification, a novel approach which leverages the alignment of complex high-order discriminative features from a mean teacher model. This alignment enhances the student model’s ability to distinguish examples across different classes. StAlK demonstrates superior performance in scenarios involving both inter and intra-class relationships and proves significantly more robust in handling class imbalance compared to baseline methods. Extensive investigations across multiple benchmark datasets reveal that StAlK achieves a substantial improvement of 6%–7% in top-1 accuracy compared to various state-of-the-art baselines. The code is available at: https://github.com/philsaurabh/StAlK_KBS.