Inter-class Similarity Research Articles

Hybrid convolutional network with enhanced graph attention mechanism (HCN-EGAM) for hyperspectral image classification

ABSTRACT Due to their reliance on superficial spectral or spatial information, current hyperspectral image classification techniques frequently fail to generate similarity measures that are accurate enough to create graph structures. We have developed a new end-to-end model, the hybrid convolutional network with enhanced graph attention mechanism (HCN-EGAM), to address this issue. It integrates an enhanced graph attention mechanism with 3D–2D hybrid convolution. In order to improve classification accuracy, this model synergistically uses the graph attention network module with the hybrid convolutional feature extraction module. The model starts by creating a 3D–2D hybrid convolutional network. This lets it quickly pull out deep spatial-spectral features that make different ground objects stand out in hyperspectral images. The model then employs these deep spatial-spectral properties to construct a graph structure, thereby enhancing its feature representation capabilities. Lastly, we use the Enhanced Graph Attention Mechanism (EGAM) to learn long-range spatial associations, successfully distinguishing between intra-class differences and inter-class similarities between distinct samples. Comparing our suggested method to other state-of-the-art methods, experimental assessments on three datasets, including the University of Pavia, Kennedy Space Center (KSC) and Indian Pines (IP) datasets, show a considerable improvement in classification accuracy.

FireNet: A Lightweight and Efficient Multi-Scenario Fire Object Detector

Fire and smoke detection technologies face challenges in complex and dynamic environments. Traditional detectors are vulnerable to background noise, lighting changes, and similar objects (e.g., clouds, steam, dust), leading to high false alarm rates. Additionally, they struggle with detecting small objects, limiting their effectiveness in early fire warnings and rapid responses. As real-time monitoring demands grow, traditional methods often fall short in smart city and drone applications. To address these issues, we propose FireNet, integrating a simplified Vision Transformer (RepViT) to enhance global feature learning while reducing computational overhead. Dynamic snake convolution (DSConv) captures fine boundary details of flames and smoke, especially in complex curved edges. A lightweight decoupled detection head optimizes classification and localization, ideal for high inter-class similarity and small targets. FireNet outperforms YOLOv8 on the Fire Scene dataset (FSD) with a mAP@0.5 of 80.2%, recall of 78.4%, and precision of 82.6%, with an inference time of 26.7 ms. It also excels on the FSD dataset, addressing current fire detection challenges.

PFEI-Net: A profound feature exploration and interaction network for ceramic substrate surface defect detection

Ceramic substrates serve as the foundational material for numerous electronic devices, and their surface quality directly affects performance and longevity. Therefore, surface defect detection on ceramic substrates is an indispensable task during the manufacturing process. Ceramic substrate surface defects exhibit low contrast, along with intraclass difference and interclass similarity, posing substantial challenges for automated visual inspection. Existing deep learning-based methods utilize carefully crafted backbone networks and feature pyramid structures that perceive multilevel information to deal with the above challenges. However, these methods still suffer from insufficient extraction of discriminative features and low effectiveness in feature fusion, thereby hindering the detection performance. To address these issues, we propose the profound feature exploration and interaction network (PFEI-Net). Initially, we propose the discriminative feature mining backbone (DFM-backbone) to progressively explore the effective discriminative features. In the DFM-backbone, the global information grated perception module is devised to construct long-range feature dependencies and aggregate rich features, whereas the contextual semantic flexible extraction module is designed to meticulously perceive contextual semantic information in a targeted manner, enhancing the comprehensive representation of the discriminative features. Then, we propose the multipath semantic interaction guided-feature pyramid network (MSIG-FPN) to facilitate the fusion and interaction of the valuable information. In the MSIG-FPN, the hierarchical refocus–reaggregation module is constructed to refocus on task-relevant features and provide reliable guidance for the network, whereas the cross-stage semantic deep fusion module is designed to integrate multi-level contextual information deeply across stages and is employed to construct the semantic complementary paths, thereby preventing the loss of crucial features. Finally, built upon the backbone of DFM-backbone and MSIG-FPN, we establish the PFEI-Net, achieving accurate detection of surface defects on ceramic substrates. Experimental results highlight that the PFEI-Net achieves excellent performance with a remarkable 89.3 % mean average precision (mAP) and an inference time of 26.1 ms, which offers a promising and viable solution for automated surface defect detection of ceramic substrate.

XAI-FruitNet: An explainable deep model for accurate fruit classification

In agricultural technology, precise fruit classification is essential yet challenging due to inherent interclass similarities and intra-class variabilities among fruit species. Despite their impressive performance, traditional deep learning models suffer from a lack of interpretability, which hampers their transparency and trustworthiness in practical applications. To address these issues, we present XAI-FruitNet, a novel hybrid deep learning architecture designed to enhance feature discrimination by integrating average and max pooling techniques. XAI-FruitNet, an optimized architecture for efficiency evaluated across the Fruits-360, Fruit Recognition, Fruit and Vegetables Image Recognition, and Dry Fruit datasets, consistently achieves over 97 % accuracy, surpassing existing state-of-the-art models and underscoring its remarkable generalization capability. A significant advancement of XAI-FruitNet is its built-in interpretability, which enhances the model's transparency and fosters trust among endusers. Through rigorous experimentation, we demonstrate that XAI-FruitNet advances state-of-the-art fruit classification accuracy and sets a new standard for explainable artificial intelligence (XAI) in agricultural applications. This hybrid approach ensures that stakeholders can rely on the classification outcomes' high performance and comprehensible nature, thereby offering a robust and trustworthy solution for modern agricultural needs.

DASR-Net: Land Cover Classification Methods for Hybrid Multiattention Multispectral High Spectral Resolution Remote Sensing Imagery

The prompt acquisition of precise land cover categorization data is indispensable for the strategic development of contemporary farming practices, especially within the realm of forestry oversight and preservation. Forests are complex ecosystems that require precise monitoring to assess their health, biodiversity, and response to environmental changes. The existing methods for classifying remotely sensed imagery often encounter challenges due to the intricate spacing of feature classes, intraclass diversity, and interclass similarity, which can lead to weak perceptual ability, insufficient feature expression, and a lack of distinction when classifying forested areas at various scales. In this study, we introduce the DASR-Net algorithm, which integrates a dual attention network (DAN) in parallel with the Residual Network (ResNet) to enhance land cover classification, specifically focusing on improving the classification of forested regions. The dual attention mechanism within DASR-Net is designed to address the complexities inherent in forested landscapes by effectively capturing multiscale semantic information. This is achieved through multiscale null attention, which allows for the detailed examination of forest structures across different scales, and channel attention, which assigns weights to each channel to enhance feature expression using an improved BSE-ResNet bilinear approach. The two-channel parallel architecture of DASR-Net is particularly adept at resolving structural differences within forested areas, thereby avoiding information loss and the excessive fusion of features that can occur with traditional methods. This results in a more discriminative classification of remote sensing imagery, which is essential for accurate forest monitoring and management. To assess the efficacy of DASR-Net, we carried out tests with 10m Sentinel-2 multispectral remote sensing images over the Heshan District, which is renowned for its varied forestry. The findings reveal that the DASR-Net algorithm attains an accuracy rate of 96.36%, outperforming classical neural network models and the transformer (ViT) model. This demonstrates the scientific robustness and promise of the DASR-Net model in assisting with automatic object recognition for precise forest classification. Furthermore, we emphasize the relevance of our proposed model to hyperspectral datasets, which are frequently utilized in agricultural and forest classification tasks. DASR-Net’s enhanced feature extraction and classification capabilities are particularly advantageous for hyperspectral data, where the rich spectral information can be effectively harnessed to differentiate between various forest types and conditions. By doing so, DASR-Net contributes to advancing remote sensing applications in forest monitoring, supporting sustainable forestry practices and environmental conservation efforts. The findings of this study have significant practical implications for urban forestry management. The DASR-Net algorithm can enhance the accuracy of forest cover classification, aiding urban planners in better understanding and monitoring the status of urban forests. This, in turn, facilitates the development of effective forest conservation and restoration strategies, promoting the sustainable development of the urban ecological environment.

Advances in Deep Learning for Skin Cancer Diagnosis

The most prevalent type of cancer worldwide is known as skin cancer. Early detection is critical because if left undiagnosed in the primary stage, it might be fatal. Although there are differences within the class and high inter-class similarities, it is too difficult to distinguish with the naked eye. Owing to the disease's global prevalence, a number of deep learning based automated systems were created thus far to help doctors identify skin lesions early on. Using pre-trained ImageNet weights and fine-tuning the Convolutional Neural Networks (CNNs), we trained VGG19 on the HAM10000 dataset. The optimal performance was observed with FT. The model that was created, which yielded an accuracy that was greater overall than the one used in transfer learning, was 82.4±1.9 %. By offering a second opinion and supporting the clinician's diagnosis, this performance could lower morbidity and treatment costs.

Remote sensing scene classification based on contextual attention mechanism of lie group space

ABSTRACT High-resolution remote sensing (HRRS) scene classification is based on the semantic information in the image and its contextual spatial correlation, labelling different semantic categories for different HRRS. Convolutional neural networks (CNN) have been widely studied and applied in remote sensing scene classification in recent years. However, most of the existing CNN models focus on the global and high-level semantic features, ignoring the shallower feature information. In addition, complex background information and variable scales lead to a series of problems with large intra-class differences and high inter-class similarities, which also brings challenges to scene classification. To address the above problems and challenges, a scenario classification model based on the contextual spatial attention and channel attention mechanism of Lie Group manifold space learning is proposed in this study. In this model, we fully explore the multi-scale features of the scene (shallower-level and high-level) and propose a novel contextual spatial attention mechanism and channel attention mechanism. Extensive experimentation was carried out on the Union Remote Sensing Image Data Set (URSIS), which improved by 7.13% compared with the classical model. The experimental results showed that compared with other state-of-the-art remote sensing scene classification models, our proposed method has achieved significant improvement in classification accuracy and performance.

Teacher-Student Prototype Enhancement Network for a Few-Shot Remote Sensing Scene Classification

Few-shot remote sensing scene classification identifies new classes from limited labeled samples where the great challenges are intraclass diversity, interclass similarity, and limited supervision. To alleviate these problems, a teacher-student prototype enhancement network is proposed for a few-shot remote sensing scene classification. Instead of introducing an attentional mechanism in mainstream studies, a prototype enhancement module is recommended to adaptively select high-confidence query samples, which can enhance the support prototype representations to emphasize intraclass and interclass relationships. The construction of a few-shot teacher model generates more discriminative predictive representations with inputs from many labeled samples, thus providing a strong supervisory signal to the student model and encouraging the network to achieve accurate classification with a limited number of labeled samples. Extensive experiments of four public datasets, including NWPU-remote sens ing image scene classification (NWPU-RESISC45), aerial image dataset (AID), UC Merced, and WHU-RS19, demonstrate that this method achieves superior competitive performance than the state-of-the-art methods on five-way, one-shot, and five-shot classifications.

Innovative multi-stage matching for counting anything

Few-shot counting (FSC) is the task of counting the number of objects in an image that belong to the same category, by using a provided exemplar pattern. By replacing the exemplar, we can effectively count anything, even in cases where we have no prior knowledge of that category’s exemplar. However, due to the variations within the same category and the impact of inter-class similarity, it is challenging to achieve accurate intra-class similarity matching using conventional similarity comparison methods. To tackle these issues, we propose a novel few-shot counting method called Multi-stage Exemplar Attention Match Network (MEAMNet), which increases the accuracy of matching, reduces the impact of noise, and enhances similarity feature matching. Specifically, we propose a multi-stage matching strategy to obtain more stable and effective matching results by acquiring similar feature in different feature spaces. In addition, we propose a novel feature matching module called Exemplar Attention Match (EAM). With this module, the intra-class similarity representation in each stage will be enhanced to achieve a better matching of the key feature. Experimental results indicate that our method not only significantly surpasses the state-of-the-art (SOTA) methods in most evaluation metrics on the FSC-147 dataset but also achieves comprehensive superiority on the CARPK dataset. This highlights the outstanding accuracy and stability of our matching performance, as well as its exceptional transferability. We will release the code at https://github.com/hzg0505/MEAMNet.

IIS-FVIQA: Finger Vein Image Quality Assessment with intra-class and inter-class similarity

In recent years, Finger Vein Image Quality Assessment (FVIQA) has been recognized as an effective solution to the problem of erroneous recognition resulting from low image quality due to false and missing information in finger vein images, and has become an important part of finger vein recognition systems. Compared to traditional FVIQA methods that rely on domain knowledge, newer methods that reject low-quality images have been favored for their independence from human interference. However, these methods only consider intra-class similarity information and ignore valuable information from inter-class distribution, which is also an important factor in evaluating the performance of recognition systems. In this work, we propose a novel FVIQA approach, named IIS-FVIQA, which concurrently takes into account the intra-class similarity density and inter-class similarity distribution distance within recognition systems. Specifically, our method generates quality scores for finger vein images by combining the information entropy of intra-class similarity distribution and Wasserstein distance of inter-class distribution. Then, we train a regression network for quality prediction using training images and corresponding quality scores. When a new image enters the recognition system, the trained regression network directly predicts the quality score of the image, making it easier for the system to select the corresponding operation based on the quality score of the image. Extensive experiments conducted on benchmark datasets demonstrate that the IIS-FVIQA method proposed in this paper consistently achieves top performance across multiple public datasets. After filtering out 10% of low-quality images predicted by the quality regression network, the recognition system’s performance improves by 43.96% (SDUMLA), 32.23% (MMCBNU_6000), and 21.20% (FV-USM), respectively. Furthermore, the method exhibits strong generalizability across different recognition algorithms (e.g., LBP, MC, and Inception V3) and datasets (e.g., SDUMLA, MMCBNU_6000, and FV-USM).

Scene Chinese Recognition with Local and Global Attention

Recently, scene Chinese recognition has attracted increasing attention. While mainstream scene text recognition methods exhibit outstanding performance in English recognition, they are considerably limited in Chinese recognition, due to inter-class similarity, intra-class variability, and complex combination of components in scene Chinese text. In this paper, we design Adaptive Position Encoding(APE) to enhance the model’s ability to perceive spatial information. Based on APE, we have innovatively designed Local Attention Module (LAM) and Global Attention Module (GAM). Specifically, LAM captures local features to identify common characteristics among characters of the same category, addressing the issue of intra-class variability. Meanwhile, LAM captures global features to identify the subordination relationships of Chinese character components. By integrating LAM and GAM, combining both local and global features, it is possible to find differences in the details among features that are fundamentally similar, thus solving the problem of inter-class similarity. Further, we contrive the transformer encoder–decoder structure to identify the vast variety of Chinese characters. Based on the Local/Global Attention Module and transformer encoder–decoder framework, we devise the novel sequence-to-sequence Local and Global Attention Network(LGANet), where both the backbone and the encoder/decoder are composed of attention mechanisms. Subsequent experiments on the Chinese scene dataset show that the recognition accuracy of our proposed LGANet is 77.3% and the normalized editing distance is 88.6%, both of which achieve the SOTA results in Fig. 1.

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.

Feature balanced re-enhanced network with multi-factor margin loss for long-tailed visual recognition

Real-world data often exhibits a long-tailed distribution, where the number of training samples for head classes far exceeds that of tail classes. This class imbalance phenomenon poses significant challenges for training deep neural networks. Existing class-aware loss methods typically focus only on the numerical relationship between class samples, blindly favoring the optimization of tail classes during the process, while neglecting the difficulty of samples and the similarity between the current class and other classes. To this end, relying only on the number relationship can easily lead to over-fitting of tail classes, thereby failing to fully utilize the potential information in the data. Therefore, we propose the Multi-Factor Margin Loss (MFMLoss), which consists of positive margin loss and negative margin loss. MFMLoss comprehensively considers three factors at three levels: overall, class, and sample: (1) quantitative relationships, (2) inter-class similarity relationships, and (3) sample recognition difficulty. The combined consideration of these three factors enables the model to pay more attention to confusing classes and difficult samples during the training process, rather than solely on tail classes, thus achieving optimization from coarse-grained to fine-grained. To further mitigate the negative impact of the imbalance between head and tail classes on feature learning, we design a new network architecture, called F-BREN. F-BREN consists of two components: the feature balancing network and the feature re-enhancement network. The former is trained with negative margin loss, which reduces the recognizability of easy samples. The latter is trained with positive margin loss, using positive margin to give more attention to hard samples, thus balancing the model’s attention to all samples. We conducted extensive experiments on four long-tailed benchmark datasets: CIFAR10-LT, CIFAR100-LT, ImageNet-LT and iNaturalist 2018, comparing the recognition accuracy of our method with eight state-of-the-art methods. The experimental results demonstrate that our proposed method outperforms the eight compared methods.

MAFDE-DN4: Improved Few-shot plant disease classification method based on Deep Nearest Neighbor Neural Network

Deep learning-based methods for accurately identifying plant diseases can be effective in improving crop yields. However, the effectiveness of these methods heavily relies on the availability of large-scale manually labeled datasets, which present technical and economic challenges. Few-shot learning methods can be generalized to new categories with a small number of samples, which is very promising in the field of plant disease recognition despite the limited sample size. However, due to the complexity of real-life scenarios, distribution of disease leaves, and significant intra-class variability and inter-class similarity resulting from crop species and disease species, existing methods perform poorly in the field of plant disease recognition. To address the above problems, this paper proposes an improved multi-scale attention fusion with discriminative enhancement deep nearest neighbor neural network MAFDE-DN4 based on DN4. Our approach makes three main contributions. First, we have designed a bidirectional weighted feature fusion module (BWFM) to enhance the aggregation of fine-grained features and enhance the network’s representation of complex disease images. Second, to tackle the issue of sparse feature descriptors being vulnerable to irrelevant noise in small sample conditions, a episodic attention module (EA) has been developed to produce scene category-relevant attention maps. This effectively mitigates the influence of irrelevant background information. Finally, we introduce additional spacing between category margins to enhance the original softmax loss function, amplify the inter-class differences to reduce the intra-class distances, and add L2 regularization constraint terms to stabilize the training process. To simulate different real-world scenarios, we set up different dataset settings. Under the 1-shot task and the 5-shot task, our method achieves 57.5% and 81.41% accuracy under the within-domain strategy and 36.54% and 51.23% accuracy under the cross-domain strategy. The experimental results show that our method outperforms existing related work in the field of plant disease recognition, whether it is a dataset with a single background or a field dataset with a complex background in a real scenario. MAFDE-DN4 based on Few-shot learning requires substantially less data on new categories of plant diseases.

GT-Net: global transformer network for multiclass brain tumor classification using MR images.

Multiclass classification of brain tumors from magnetic resonance (MR) images is challenging due to high inter-class similarities. To this end, convolution neural networks (CNN) have been widely adopted in recent studies. However, conventional CNN architectures fail to capture the small lesion patterns of brain tumors. To tackle this issue, in this paper, we propose a global transformer network dubbed GT-Net for multiclass brain tumor classification. The GT-Net mainly comprises a global transformer module (GTM), which is introduced on the top of a backbone network. A generalized self-attention block (GSB) is proposed to capture the feature inter-dependencies not only across spatial dimension but also channel dimension, thereby facilitating the extraction of the detailed tumor lesion information while ignoring less important information. Further, multiple GSB heads are used in GTM to leverage global feature dependencies. We evaluate our GT-Net on a benchmark dataset by adopting several backbone networks, and the results demonstrate the effectiveness of GTM. Further, comparison with state-of-the-art methods validates the superiority of our model.

An explainable MHSA enabled deep architecture with dual-scale convolutions for methane source classification using remote sensing

Methane is the second most abundant greenhouse gas after carbon dioxide. Anthropogenic sources are the dominant emitters of methane. The poor spatial resolution of satellite imagery, high interclass similarity, the multi-scalar nature of features, and the dominance of background limit the performance of the previous approaches. Further, the reliance on high-resolution imagery limits the cost-effective global application of the works introduced in the literature. To resolve this, the present work proposes a novel method for methane source classification based on open-source multi-spectral satellite imagery of Sentinel-1 and 2. The work utilizes deep dual-scale convolutions with scaled dot product self-attention calculated across the 15 composite bands of Sentinel-1 and 2 data. The incorporation of non-RGB bands along with the RGB bands further enables the model to learn the spectral differences essential for the classification. The experimental results witness the superior performance of the developed method against other considered state-of-the-art methods.

A lightweight self-supervised learning segmentation model for variable and complex high-resolution remote sensing images

The complexity and variability of high-resolution remote sensing data, such as high intra-class variability and inter-class similarity, pose significant challenges to model segmentation. To address the problem, this paper constructs a lightweight self-supervised learning model for multi-label segmentation in the form of phased learning of multi-scale features. The model adopts axial depthwise separable convolutions to reduce computational complexity and enhance feature representation, utilizes dilated rates to capture large-scale and multi-scale contextual information for long-distance feature extraction, and incorporates convolution kernels of varying sizes to acquire both local and global feature information for the improved ability of learning feature representation. The experimental results show that our model achieves competitive performance and has smaller weight parameters, memory usage, and lower computational complexity compared with existing classical models that rely on large-scale weight parameters. Additionally, our ablation study delves into the encountered design issues to elucidate the rationality of our approach. The source code is avaiable: https://github.com/zhangjy2008327/remote-sensing-images.

A mutual reconstruction network model for few-shot classification of histological images: addressing interclass similarity and intraclass diversity.

The automated classification of histological images is crucial for the diagnosis of cancer. The limited availability of well-annotated datasets, especially for rare cancers, poses a significant challenge for deep learning methods due to the small number of relevant images. This has led to the development of few-shot learning approaches, which bear considerable clinical importance, as they are designed to overcome the challenges of data scarcity in deep learning for histological image classification. Traditional methods often ignore the challenges of intraclass diversity and interclass similarities in histological images. To address this, we propose a novel mutual reconstruction network model, aimed at meeting these challenges and improving the few-shot classification performance of histological images. The key to our approach is the extraction of subtle and discriminative features. We introduce a feature enhancement module (FEM) and a mutual reconstruction module to increase differences between classes while reducing variance within classes. First, we extract features of support and query images using a feature extractor. These features are then processed by the FEM, which uses a self-attention mechanism for self-reconstruction of features, enhancing the learning of detailed features. These enhanced features are then input into the mutual reconstruction module. This module uses enhanced support features to reconstruct enhanced query features and vice versa. The classification of query samples is based on weighted calculations of the distances between query features and reconstructed query features and between support features and reconstructed support features. We extensively evaluated our model using a specially created few-shot histological image dataset. The results showed that in a 5-way 10-shot setup, our model achieved an impressive accuracy of 92.09%. This is a 23.59% improvement in accuracy compared to the model-agnostic meta-learning (MAML) method, which does not focus on fine-grained attributes. In the more challenging, 5-way 1-shot setting, our model also performed well, demonstrating a 18.52% improvement over the ProtoNet, which does not address this challenge. Additional ablation studies indicated the effectiveness and complementary nature of each module and confirmed our method's ability to parse small differences between classes and large variations within classes in histological images. These findings strongly support the superiority of our proposed method in the few-shot classification of histological images. The mutual reconstruction network provides outstanding performance in the few-shot classification of histological images, successfully overcoming the challenges of similarities between classes and diversity within classes. This marks a significant advancement in the automated classification of histological images.

EfficientNet-ECA: A lightweight network based on efficient channel attention for class-imbalanced welding defects classification

Welding defects recognition is crucial for ensuring weldment quality in robot arc welding. Nevertheless, due to unbalanced data distributions, limited computing resources in factory, as well as intraclass variability and interclass similarity among different welding defects, it is difficult to extract the most discriminative defect features from welding molten pool images on site, resulting in weak class-imbalanced defect recognition performance. To address the above issue, a novel lightweight network named EfficientNet-ECA is proposed for recognizing and classifying welding defects according to molten pool images of robot arc welding. Firstly, the EfficientNet-ECA network based on Efficient Channel Attention (ECA) is designed to extract the most discriminative features of different defects from molten pool images, where ECA is employed to enhance cross-channel information interactions. Secondly, a dynamic equalized focal loss function is proposed to both rebalance the loss contribution of class-imbalanced data of different defects and improve defects recognition accuracy. Subsequently, a class-imbalanced dataset containing eight typical welding defects is constructed to evaluate the EfficientNet-ECA model. Finally, the proposed model is comprehensively analyzed through ablation studies and compared with the existing state-of-the-art lightweight models, and results show that the proposed method exhibits better effectiveness and generalizability in classifying class-imbalanced defects across different welding scenarios, achieving the highest accuracy of 95.84% on a self-constructed AL5083 dataset and 96.50% on a publicly available SS304 dataset. Moreover, with fewer parameters and less complexity, the proposed method is more suitable for online welding quality monitoring in factory.

Pest-ConFormer: A hybrid CNN-Transformer architecture for large-scale multi-class crop pest recognition

Crop pests are acknowledged as the major factors in reducing the yield and quality of agricultural production worldwide. It is an urgent necessity to recognize crop pests accurately to protect the crop in the early stage to reduce the loss for the agricultural economy. Due to the ecological characteristics of the crop pests and the complex background in fields, the crop pests show high inter-class similarity and significant intra-class variation in external morphology appearance, which makes current recognition methods suffer from low classification accuracy and poor generalization ability in complex natural environment recognition tasks. To tackle this problem, a hybrid convolutional neural network and transformer-based model, namely Pest-ConFormer, featured with multi-scale weakly supervised feature selection mechanisms is proposed, which has shown excellent multi-scale discriminative feature extraction in fine-grained visual classification (FGVC) tasks. This method employs a hybrid convolution-transformer encoder architecture pre-training in a self-supervised masked autoencoder manner as a backbone to learn pests’ highly discriminative features across various scales. Next, a dual-path feature aggregation structure with a top-down FPN-like feature pathway and a bottom-up PANet-like feature pathway based on attention mechanisms is designed to learn high-level global context information and low-level local detailed feature representation. Thirdly, a fine-grained classification module using weakly supervised learning is introduced to select the discriminative feature points in different pyramidal levels. Then, these feature points are fed into a graph convolutional network to accomplish classification. Several experiments are conducted on the large-scale multi-class IP102 benchmark dataset, and the proposed method achieves an accuracy of 77.81 % regarding crop pest recognition. The experimental results indicate that our approach outperforms other state-of-the-art methods by nearly 2 percent points, demonstrating that the proposed hybrid architecture with dual-path feature aggregation and fine-grained classification modules can be more effective in the crop pest recognition field than CNN-based methods and can be deployed in the practical natural environment. The source code will be available at https://github.com/mwfang/pestconformer.