Inter-class Discrimination Research Articles

Robotic MAG welding defects and quality assessment with a defect threshold decision model-driven method

The use of machine vision and deep learning methods for online monitoring and evaluation of welding quality during the welding process and timely detection and correction of welding defects is essential for intelligent welding manufacturing. In this paper, with the goal of online monitoring of the welding quality of the robotic arc welding for the excavator boom, we established a multi-source sensing system for robotic arc welding processes. We have developed various feature extraction algorithms for extracting information about welding pools, arc sounds, currents, voltages, and other features. We proposed to extract welding pool image features using Res2-MobileNetV3 to obtain a 12-dimensional welding pool feature, which enabled good inter-class discrimination among different welding defects. We designed a welding defect threshold decision (DD) strategy model using the responsive machine learning and lightweight network Res2-MobileNetV3 method, achieving a welding defect recognition accuracy of 96.59%. Finally, we tested the accuracy and reliability of the welding defect recognition model on an actual welding scene for the robotic welding of the excavator boom. It provided valuable scientific methods and technological approaches for intelligent welding in complex scenes.

Joint Weakly Supervised Image Emotion Analysis Based on Interclass Discrimination and Intraclass Correlation

Joint Weakly Supervised Image Emotion Analysis Based on Interclass Discrimination and Intraclass Correlation

Deep conditional adversarial subdomain adaptation network for unsupervised mechanical fault diagnosis

In recent years, intelligent fault diagnosis technology has rapidly advanced, achieving remarkable results. Most of them assume that the source and target domains have similar distributions. However, the actual working conditions of rotating machinery are variable, leading to a significant distribution discrepancy between the source and target domains. The majority of existing domain adaptation methods mainly consider either the marginal distribution or the conditional distribution between the source and target domains, which limits the improvement of diagnostic performance in the target domain. Considering the discrepancy in domain distributions, a deep conditional adversarial subdomain adaptation network (DCASAN) is proposed in this paper for unsupervised mechanical fault diagnosis under unknown working conditions. In DCASAN, a backbone network based on the feature fusion of convolutional neural network (CNN) and graph convolutional network (GCN) is designed for feature extraction. Furthermore, a conditional adversarial subdomain adaptation strategy is proposed, which simultaneously considers both the marginal distribution and the conditional distribution. Under the joint constraints of marginal and conditional distribution alignment, the extracted features exhibit improved inter-class discriminability and intra-class consistency, effectively reducing the distribution discrepancy between different domains. Results from multiple transfer tasks on three rotating machinery datasets demonstrate that the proposed method can learn rich transferable features. The overall average accuracy of the proposed method on these three datasets is 88.07%, 89.17%, and 96.63%, respectively, significantly outperforming other methods in terms of robustness and generalization.

Deep discriminative sparse representation learning for machinery fault diagnosis

The high complexity of actual machinery vibration environments introduces various interferences into vibration signals, making it challenging to eliminate redundant information and extract discriminative features of mechanical faults. Shallow-structured sparse classification models enable intelligent fault diagnosis but struggle with deeper feature extraction, especially in small-sample cases. In this paper, we introduce a deep discriminative sparse representation learning (DDSRL) framework with a deep architecture for this issue. In DDSRL, the first-layer dictionary atoms are automatically learned from raw signals and borrowed for sparse coding to enhance sparsity and interclass discriminability. Then, the weight matrix is defined to screen high-energy atoms from the upper dictionary as new training samples for the next dictionary learning and sparse coding layers, thereby eliminating interclass redundant atoms and mining deeper discriminative representations of cyclostationary impulses or meshing harmonics from the shallow dictionary. Finally, DDSRL extracts feature sequences sensitive to fault-induced waveforms via the learned dictionaries and coding coefficients for intelligent fault identification. Experimental verifications on two datasets with high-speed bearing and compound gear-bearing faults demonstrate that DDSRL outperforms six popular sparse classification models, sparse autoencoder, convolutional neural network, and principal component analysis network. Furthermore, the layer-by-layer feature extraction process of DDSRL and feature visualization analysis provide more credible fault diagnosis results.

Few-Shot Object Detection in Remote Sensing Images via Data Clearing and Stationary Meta-Learning

Nowadays, the focus on few-shot object detection (FSOD) is fueled by limited remote sensing data availability. In view of various challenges posed by remote sensing images (RSIs) and FSOD, we propose a meta-learning-based Balanced Few-Shot Object Detector (B-FSDet), built upon YOLOv9 (GELAN-C version). Firstly, addressing the problem of incompletely annotated objects that potentially breaks the balance of the few-shot principle, we propose a straightforward yet efficient data clearing strategy, which ensures balanced input of each category. Additionally, considering the significant variance fluctuations in output feature vectors from the support set that lead to reduced effectiveness in accurately representing object information for each class, we propose a stationary feature extraction module and corresponding stationary and fast prediction method, forming a stationary meta-learning mode. In the end, in consideration of the issue of minimal inter-class differences in RSIs, we propose inter-class discrimination support loss based on the stationary meta-learning mode to ensure the information provided for each class from the support set is balanced and easier to distinguish. Our proposed detector’s performance is evaluated on the DIOR and NWPU VHR-10.v2 datasets, and comparative analysis with state-of-the-art detectors reveals promising performance.

Two-stage residual networks for damage identification and location of stiffened composite panel based on guided waves

The damage detection of the stiffened composite panel, as a typical aircraft structure, is a research hotspot in Structural Health Monitoring (SHM). where guided waves propagate with multi-modal and dispersion characteristics. The traditional damage detection method manually extracts the potential discriminative features of the signal to achieve damage identification, depending on expert experience. In this paper, we propose a two-stage residual networks (ResNets) framework based on guided waves to locate damage in the stiffened composite panel, which automatically mines the high-dimensional features with sensitive discriminant information. The guided wave signal acquisition system collects four types of data: health data, stringer damage data, damage data on the skin of the stringer-side, and damage data on the skin-side. The first-stage utilizes a ResNet to classify the structure condition, while in the second-stage, three separate ResNets are employed to locate the damage according to the classification results of the first-stage. The experimental results show that the accuracy of the first-stage damage classification and the damage localization of the stringer and the skin of the stringer-side in the second-stage has reached 100%, and that of the skin-side is 99.13%, which significantly outperforms single-stage methods. This strategy of inter-class discrimination and intra-class precise localization of damage can not only identify the damaged regions but also determine the specific location of the damage, which greatly increases the performance of SHM. The present two-stage method is a potential solution for future SHM strategies and further investigation is warranted.

Class-specific regularized joint distribution alignment for unsupervised domain adaptation

Unsupervised domain adaptation (UDA) aims to learn robust classifiers for the target domain by leveraging knowledge from annotated source domain. Existing methods concentrated on cross-domain distribution alignment or domain-invariant deep learning model construction. Recently, although the class-specific characteristics have been incorporated for numerous UDA methods, there remains a need for further exploration in capturing intra-class and inter-class characteristics between the source and target domains. In this paper, we propose a novel UDA method, referred to as Class-specific Regularized Joint Distribution Alignment (CRJDA), to simultaneously optimize the intra-class distance of source domain, the inter-class discriminability of target domain, and the joint distribution between domains. Specifically, our method involves an overall optimization process that minimizes the maximum mean discrepancy while incorporating dual penalized class-specific regularizations during joint distribution alignment. Extensive experiments conducted on several benchmark datasets demonstrate that the superiority of the proposed method compared to both state-of-the-art conventional UDA methods and advanced deep UDA models.

Addressing Skewed Heterogeneity via Federated Prototype Rectification With Personalization.

Federated learning (FL) is an efficient framework designed to facilitate collaborative model training across multiple distributed devices while preserving user data privacy. A significant challenge of FL is data-level heterogeneity, i.e., skewed or long-tailed distribution of private data. Although various methods have been proposed to address this challenge, most of them assume that the underlying global data are uniformly distributed across all clients. This article investigates data-level heterogeneity FL with a brief review and redefines a more practical and challenging setting called skewed heterogeneous FL (SHFL). Accordingly, we propose a novel federated prototype rectification with personalization (FedPRP) which consists of two parts: federated personalization and federated prototype rectification. The former aims to construct balanced decision boundaries between dominant and minority classes based on private data, while the latter exploits both interclass discrimination and intraclass consistency to rectify empirical prototypes. Experiments on three popular benchmarks show that the proposed approach outperforms current state-of-the-art methods and achieves balanced performance in both personalization and generalization.

High Precision Cervical Precancerous Lesion Classification Method Based on ConvNeXt.

Traditional cervical cancer diagnosis mainly relies on human papillomavirus (HPV) concentration testing. Considering that HPV concentrations vary from individual to individual and fluctuate over time, this method requires multiple tests, leading to high costs. Recently, some scholars have focused on the method of cervical cytology for diagnosis. However, cervical cancer cells have complex textural characteristics and small differences between different cell subtypes, which brings great challenges for high-precision screening of cervical cancer. In this paper, we propose a high-precision cervical cancer precancerous lesion screening classification method based on ConvNeXt, utilizing self-supervised data augmentation and ensemble learning strategies to achieve cervical cancer cell feature extraction and inter-class discrimination, respectively. We used the Deep Cervical Cytological Levels (DCCL) dataset, which includes 1167 cervical cytology specimens from participants aged 32 to 67, for algorithm training and validation. We tested our method on the DCCL dataset, and the final classification accuracy was 8.85% higher than that of previous advanced models, which means that our method has significant advantages compared to other advanced methods.

SSL-ProtoNet: Self-supervised Learning Prototypical Networks for few-shot learning

Few-shot learning is seeking to generalize well to unseen tasks with insufficient labeled samples. Existing works have achieved generalization by exploring inter-class discrimination. However, their performance is limited because sample discrimination is neglected. In this work, we propose a metric-based few-shot approach that leverages self-supervised learning, Prototypical networks, and knowledge distillation, referred to as SSL-ProtoNet, to utilize sample discrimination. The proposed SSL-ProtoNet consists of three stages: pre-training stage, fine-tuning stage, and self-distillation stage. In the pre-training stage, self-supervised learning is leveraged to cluster the samples with their augmented variants to enhance the sample discrimination. The learned representation is then served as an initial point for the next stage. In the fine-tuning stage, the model weights transferred from the pre-training stage are fine-tuned to the target few-shot tasks. A self-supervised loss and a few-shot loss are integrated to prevent overfitting during few-shot task adaptation and to maintain the embedding diversity. In the self-distillation stage, the model is arranged in a teacher–student architecture. The teacher model will serve as a guidance in student model training to reduce overfitting and further improve the performance. The experimental results show that the proposed SSL-ProtoNet outshines the state-of-the-art few-shot image classification methods on three benchmark few-shot datasets, namely, miniImageNet, tieredImageNet, and CIFAR-FS. The source code for the proposed method is available at https://github.com/Jityan/sslprotonet.

Electroencephalograph Emotion Classification Using a Novel Adaptive Ensemble Classifier Considering Personality Traits.

The study explores the use of Electroencephalograph (EEG) signals as a means to uncover various states of the human brain, with a specific focus on emotion classification. Despite the potential of EEG signals in this domain, existing methods face challenges. Features extracted from EEG signals may not accurately represent an individual's emotional patterns due to interference from time-varying factors and noise. Additionally, higher-level cognitive factors, such as personality, mood, and past experiences, further complicate emotion recognition. The dynamic nature of EEG data in terms of time series introduces variability in feature distribution and interclass discrimination across different time stages. To address these challenges, the paper proposes a novel adaptive ensemble classification method. The study introduces a new method for providing emotional stimuli, categorizing them into three groups (sadness, neutral, and happiness) based on their valence-arousal (VA) scores. The experiment involved 60 participants aged 19-30 years, and the proposed method aimed to mitigate the limitations associated with conventional classifiers. The results demonstrate a significant improvement in the performance of emotion classifiers compared to conventional methods. The classification accuracy achieved by the proposed adaptive ensemble classification method is reported at 87.96%. This suggests a promising advancement in the ability to accurately classify emotions using EEG signals, overcoming the limitations outlined in the introduction. In conclusion, the paper introduces an innovative approach to emotion classification based on EEG signals, addressing key challenges associated with existing methods. By employing a new adaptive ensemble classification method and refining the process of providing emotional stimuli, the study achieves a noteworthy improvement in classification accuracy. This advancement is crucial for enhancing our understanding of the complexities of emotion recognition through EEG signals, paving the way for more effective applications in fields such as neuroinformatics and affective computing.

Neural network representations for the inter- and intra-class common vector classifiers

Common Vector Approach (CVA) is a known linear regression-based classifier, which also enables an extension to inter-class discrimination, known as the Discriminative Common Vector Approach (DCVA). The characteristics of linear regression classifiers (LRCs) enable the possibility of a schematic implementation that is similar to the neuron model of artificial neural networks (ANNs). In this work, we explore this schematic similarity to come up with an ANN representation for both CVA and DCVA. The new representation eliminates the need for projection matrices in its implementation, hence significantly reduces the memory requirements and computational complexities of the processes. Furthermore, since the new representation is in a neural style, it is expected to provide a solid and intriguing extension of CVA (and DCVA) by further incorporating adaptation or activation processes to the already successful CVA-based classifiers.

Task-Adaptive Feature Disentanglement and Hallucination for Few-Shot Classification

Few-shot classification is a challenging task of computer vision and is critical to the data-sparse scenario like rare disease diagnosis. Feature augmentation is a straightforward way to alleviate the data-sparse issue in few-shot classification. However, mimicking the original feature distribution from a small amount of data is challenging. Existing augmentation-based methods are task-agnostic: the augmented feature is not with optimal intra-class diversity and inter-class discriminability concerning a certain task. To address this drawback, we propose a novel <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Task-adaptive Feature Disentanglement and Hallucination</i> framework, dubbed TaFDH. Concretely, we first perceive the task information to disentangle the original feature into two components: class-irrelevant and class-specific features. Then more class-irrelevant features are decoded from a learned variational distribution, fused with the class-specific feature to get the augmented features. Finally, a generalized prior distribution over a quadratic classifier is meta-learned, which can be fast adapted to the class-specific posterior, thus further alleviating the inadequacy and uncertainty of feature hallucination via the nature of Bayesian inference. In this way, we construct a more discriminable embedding space with reasonable intra-class diversity instead of simply restoring the original embedding space, which can lead to a more precise decision boundary. We obtain the augmented features equipped with enhanced inter-class discriminability by highlighting the most discriminable part while boosting the intra-class diversity by fusing with the diverse generated class-irrelevant parts. Experiments on five multi-grained few-shot classification datasets demonstrate the superiority of our method.

Content-based medical image retrieval with opponent class adaptive margin loss

The increasing utilization of medical imaging technology with digital storage capabilities has facilitated the compilation of large-scale data repositories. Fast access to image samples with similar appearance to suspected cases in these repositories can help establish a consulting system for healthcare professionals, and improve diagnostic procedures while minimizing processing delays. However, manual querying of large repositories is labor intensive. Content-based image retrieval (CBIR) offers an automated solution based on quantitative assessment of image similarity based on image features in a latent space. Since conventional methods based on hand-crafted features typically show poor generalization performance, learning-based CBIR methods have received attention recently. A common framework in this domain involves classifier-guided models that are trained to detect different image classes. Similarity assessments are then performed on the features captured by the intermediate stages of the trained models. While classifier-guided methods are powerful in inter-class discrimination, they are suboptimally sensitive to within-class differences in image features. An alternative framework instead performs task-agnostic training to learn an embedding space that enforces the representational discriminability of images. Within this representational-learning framework, a powerful method is triplet-wise learning that addresses the deficiencies of point-wise and pair-wise learning in characterizing the similarity relationships between image classes. However, the traditional triplet loss enforces separation between only a subset of image samples within the triplet via a manually-set constant margin value, so it can lead to suboptimal segregation of opponent classes and limited generalization performance. To address these limitations, we introduce a triplet-learning method for automated querying of medical image repositories based on a novel Opponent Class Adaptive Margin (OCAM) loss. To maintain optimally discriminative representations, OCAM considers relationships among all image pairs within the triplet and utilizes an adaptive margin value that is automatically selected per dataset and during the course of training iterations. CBIR performance of OCAM is compared against state-of-the-art loss functions for representational learning on three public databases (gastrointestinal disease, skin lesion, lung disease). On average, OCAM shows an mAP performance of 86.30% in the KVASIR dataset, 70.30% in the ISIC 2019 dataset, and 85.57% in the X-RAY dataset. Comprehensive experiments in each application domain demonstrate the superior performance of OCAM against competing triplet-wise methods at 1.52%, classifier-guided methods at 2.29%, and non-triplet representational-learning methods at 4.56%.

Intra-class consistency and inter-class discrimination feature learning for automatic skin lesion classification.

Automated skin lesion classification has been proved to be capable of improving the diagnostic performance for dermoscopic images. Although many successes have been achieved, accurate classification remains challenging due to the significant intra-class variation and inter-class similarity. In this article, a deep learning method is proposed to increase the intra-class consistency as well as the inter-class discrimination of learned features in the automatic skin lesion classification. To enhance the inter-class discriminative feature learning, a CAM-based (class activation mapping) global-lesion localization module is proposed by optimizing the distance of CAMs for the same dermoscopic image generated by different skin lesion tasks. Then, a global features guided intra-class similarity learning module is proposed to generate the class center according to the deep features of all samples in one class and the history feature of one sample during the learning process. In this way, the performance can be improved with the collaboration of CAM-based inter-class feature discriminating and global features guided intra-class feature concentrating. To evaluate the effectiveness of the proposed method, extensive experiments are conducted on the ISIC-2017 and ISIC-2018 datasets. Experimental results with different backbones have demonstrated that the proposed method has good generalizability and can adaptively focus on more discriminative regions of the skin lesion.

Real-time grouping of tobacco through channel weighting and dynamic loss regulation

Grouping of tobacco is a key technology in intelligent management of tobacco planting and fine blending in cigarette industry. However, existing grouping methods based on deep learning are limited by lacking key feature expression due to excessive abstraction at higher scales. In addition, the inter-class discrimination ability is limited and the model prefers to majority classes during training. To tackle these problems, we propose a real-time tobacco grouping network (TGNet) based on channel weighting of higher scale features and dynamic loss regulation. Firstly, the industrial array camera equipped on a machine vision detection platform is used to capture the original tobacco images, which including the orange (F), lemon (L), variegated (K), greenish (V), and green-yellow (GY) categories manually labeled by grading experts. Then, to tackle the deficiency of key feature expression in the higher scale abstraction layers, a feature channel weighting module is proposed, which generates weighting factors proportional to different channels to re-encode higher scale features and emphatically highlight the key area and color information. The lightweight structure of feature channel weighting module compresses the model volume and improves forward inference speed. Finally, aiming at the problems of inter-group ambiguity and class imbalance dilemma, the focal and dynamic margin (FADM) loss with focusing on minority classes training and adaptative margin is proposed. The dynamic margin is adaptatively encoded by a Cosine function of similarity, which uses tobacco category recognition probability to measure the similarity of tobacco. Extensive performance evaluations show that TGNet outperforms other classification models in terms of the average classification accuracy 77.16%, inference time 112FPS, model size 2.5MB, model computation 0.12GFlops and model parameter number 0.613 M. The classification accuracy of minority classes V and GY is improved.

Reduced volume of diabetic pancreatic islets in rodents detected by synchrotron X-ray phase-contrast microtomography and deep learning network

The pancreatic islet is a highly structured micro-organ that produces insulin in response to rising blood glucose. Here we develop a label-free and automatic imaging approach to visualize the islets in situ in diabetic rodents by the synchrotron radiation X-ray phase-contrast microtomography (SRμCT) at the ID17 station of the European Synchrotron Radiation Facility. The large-size images (3.2mm×15.97mm) were acquired in the pancreas in STZ-treated mice and diabetic GK rats. Each pancreas was dissected by 3000 reconstructed images. The image datasets were further analysed by a self-developed deep learning method, AA-Net. All islets in the pancreas were segmented and visualized by the three-dimension (3D) reconstruction. After quantifying the volumes of the islets, we found that the number of larger islets (=>1500 μm3) was reduced by 2-fold (wt 1004±94 vs GK 419±122, P<0.001) in chronically developed diabetic GK rat, while in STZ-treated diabetic mouse the large islets were decreased by half (189±33 vs 90±29, P<0.001) compared to the untreated mice. Our study provides a label-free tool for detecting and quantifying pancreatic islets in situ. It implies the possibility of monitoring the state of pancreatic islets in vivo diabetes without labelling.

Vehicle Tracking Algorithm Based on Deep Learning in Roadside Perspective

Traffic intelligence has become an important part of the development of various countries and the automobile industry. Roadside perception is an important part of the intelligent transportation system, which mainly realizes the effective perception of road environment information by using sensors installed on the roadside. Vehicles are the main road targets in most traffic scenes, so tracking a large number of vehicles is an important subject in the field of roadside perception. Considering the characteristics of vehicle-like rigid targets from the roadside view, a vehicle tracking algorithm based on deep learning was proposed. Firstly, we optimized a DLA-34 network and designed a block-N module, then the channel attention and spatial attention modules were added in the front of the network to improve the overall feature extraction ability and computing efficiency of the network. Next, the joint loss function was designed to improve the intra-class and inter-class discrimination ability of the tracking algorithm, which can better discriminate objects of similar appearance and the color of vehicles, alleviate the IDs problem and improve algorithm robustness and the real-time performance of the tracking algorithm. Finally, the experimental results showed that the method had a good tracking effect for the vehicle tracking task from the roadside perspective and could meet the practical application demands of complex traffic scenes.

Spectral–Spatial Masked Transformer With Supervised and Contrastive Learning for Hyperspectral Image Classification

Recently, due to the powerful capability at modeling the long-range relationships, Transformer-based methods have been widely explored in many research areas including hyperspectral image (HSI) classification. However, because of lots of trainable parameters and the lack of inductive bias, it is difficult to train a Transformer-based HSI classifier, especially when the number of training samples is limited. To address this issue, in this study, spectral-spatial masked Transformer (SS-MTr) is explored for HSI classification, which uses a two-stage training strategy. In the first stage, SS-MTr pre-trains a vanilla Transformer via reconstruction from masked HSI inputs, which embeds the local inductive bias into the Transformer. In the second stage, the well pre-trained Transformer is cooperated with a fully connected layer and then fine-tuned for the HSI classification. Furthermore, in order to incorporate discriminative feature learning into the SS-MTr, three SS-MTr-based methods, including contrastive SS-MTr (C-SS-MTr), supervised SS-MTr (S-SS-MTr), and supervised contrastive SS-MTr (SC-SS-MTr) are proposed by adding extra branches for specific tasks in parallel with the existing reconstruction task. Specifically, the proposed C-SS-MTr adds a contrastive loss which brings instance discriminability. Besides, the proposed S-SS-MTr builds an extra classification branch for embracing inter-class discriminability and intra-class similarity. Moreover, the proposed SC-SS-MTr combines C-SS-MTr and S-SS-MTr for better generalization. The proposed SS-MTr, C-SS-MTr, S-SS-MTr, and SC-SS-MTr are tested on three popular hyperspectral datasets (i.e., Indian Pines, Pavia University, and Houston). The obtained results reveal that the proposed models achieve competitive results compared with the state-of-the-art HSI classification methods. Code is available at https://github.com/mengduanjinghua/SS-MTr.

M2FN: An end-to-end multi-task and multi-sensor fusion network for intelligent fault diagnosis

Intelligent fault diagnosis based on multi-sensor fusion has gained considerable attention in various modern industrial applications. However, it is still challenging to extract discriminative features from multi-sensor data to provide an accurate and reliable diagnosis. For this purpose, this paper proposes a new multi-task multi-sensor fusion network (M2FN) to improve fault diagnosis performance. The proposed method first uses convolutional neural networks to extract and fuse features from raw vibration and current signals. After that, to improve the discriminative ability of the learned features, a multi-task learning module (MTL) is designed which contains a classification task and a deep metric learning task. Our proposed M2FN model is evaluated on a bearing dataset and a gearbox dataset. Experimental results show that our proposed M2FN method significantly outperforms the compared single-sensor-based and single-task-based methods in terms of diagnosis accuracy, and the learned features present better inter-class discriminability and intra-class concentration through the feature visualization analysis.