Split Network Research Articles

Multi-scale receptive field grouped and split attention network for real-time detection of hazardous chemical gas leaks in infrared images

Abstract The petrochemical industry faces frequent hazardous gas leaks, which demand precise and timely detection to avert severe consequences. Existing computer vision approaches encounter challenges due to limitations in gas characteristics and scene features. To address these issues, we propose a multiscale receptive field grouped and split attention network, GAS-YOLO, that integrates infrared imaging technology. Within GAS-YOLO, we design a novel module, multi-scale receptive field grouped convolution (MRFGConv), to preserve fine-grained information, preventing detail loss and addressing spatial attention feature-sharing issues. An innovative split convolution attention (SCA) mechanism in the C2f module effectively couples multi-scale features, balancing performance and efficiency. Additionally, the asymptotic feature pyramid network (AFPN) facilitates the mutual interaction of information between non-adjacent levels, enabling advanced feature fusion. Using benchmark InfraGasLeakDataset, GAS-YOLO surpasses YOLOv8-n by 5.8% mAP50, with SCA outperforming state-of-the-art attention models. Experiment results validate the effectiveness and feasibility of our proposed approaches, providing valuable insights into hazardous chemical gas leak detection.

A detection network for small defects of steel surface based on YOLOv7

Defect detection is crucial in steel production to prevent safety risks and prolong lifespan. However, current methods still face challenges in accurately detecting small and vague targets. In response to this problem, we propose a model named SRN-YOLO for steel surface detection based on YOLOv7. Firstly, a split residual convolution network (SResNet) is designed to capture gradient feature information. Then, a re-fusion feature pyramid network (RFPN) is built to minimize the loss of features. Furthermore, a location regression loss function named NWD-CIOU combined the Normalized Wasserstein Distance (NWD) and Complete Intersection Over Union (CIOU) is employed to increase the number of positive samples of small targets. The experiment results show that SRN-YOLO has excellent detection performance for steel surface defects. Specifically, the recognition accuracies of the model on NEU-DET and GC10-DET reach 81.2 mAP and 71.6 mAP, respectively, which are improved by 5.8% and 3.8% compared to YOLOv7.

Multi-Level Attention Split Network: A Novel Malaria Cell Detection Algorithm

Malaria is one of the major global health threats. Microscopic examination has been designated as the “gold standard” for malaria detection by the World Health Organization. However, it heavily relies on the experience of doctors, resulting in long diagnosis time, low efficiency, and a high risk of missed or misdiagnosed cases. To alleviate the pressure on healthcare workers and achieve automated malaria detection, numerous target detection models have been applied to the blood smear examination for malaria cells. This paper introduces the multi-level attention split network (MAS-Net) that improves the overall detection performance by addressing the issues of information loss for small targets and mismatch between the detection receptive field and target size. Therefore, we propose the split contextual attention structure (SPCot), which fully utilizes contextual information and avoids excessive channel compression operations, reducing information loss and improving the overall detection performance of malaria cells. In the shallow detection layer, we introduce the multi-scale receptive field detection head (MRFH), which better matches targets of different scales and provides a better detection receptive field, thus enhancing the performance of malaria cell detection. On the NLM—Malaria Dataset provided by the National Institutes of Health, the improved model achieves an average accuracy of 75.9% in the public dataset of Plasmodium vivax (malaria)-infected human blood smear. Considering the practical application of the model, we introduce the Performance-aware Approximation of Global Channel Pruning (PAGCP) to compress the model size while sacrificing a small amount of accuracy. Compared to other state-of-the-art (SOTA) methods, the proposed MAS-Net achieves competitive results.

Dynamic feature distillation and pyramid split large kernel attention network for lightweight image super-resolution

Dynamic feature distillation and pyramid split large kernel attention network for lightweight image super-resolution

A neural machine translation method based on split graph convolutional self-attention encoding

With the continuous advancement of deep learning technologies, neural machine translation (NMT) has emerged as a powerful tool for enhancing communication efficiency among the members of cross-language collaborative teams. Among the various available approaches, leveraging syntactic dependency relations to achieve enhanced translation performance has become a pivotal research direction. However, current studies often lack in-depth considerations of non-Euclidean spaces when exploring interword correlations and fail to effectively address the model complexity arising from dependency relation encoding. To address these issues, we propose a novel approach based on split graph convolutional self-attention encoding (SGSE), aiming to more comprehensively utilize syntactic dependency relationships while reducing model complexity. Specifically, we initially extract syntactic dependency relations from the source language and construct a syntax dependency graph in a non-Euclidean space. Subsequently, we devise split self-attention networks and syntactic semantic self-attention networks, integrating them into a unified model. Through experiments conducted on multiple standard datasets as well as datasets encompassing scenarios related to team collaboration and enterprise management, the proposed method significantly enhances the translation performance of the utilized model while effectively mitigating model complexity. This approach has the potential to effectively enhance communication among cross-language team members, thereby ameliorating collaborative efficiency.

Wind power forecasting: A transfer learning approach incorporating temporal convolution and adversarial training

Accurate forecasting of wind power is crucial for efficiently scheduling power grids that incorporate wind energy. However, distribution instability of time-varying data can undermine the generalization capability of forecasting models in future periods. This paper introduces a transfer learning approach incorporating adversarial training and temporal convolution to address wind power forecasting from the distribution-centric perspective. The proposed approach can be divided into two joint modules: the temporal domain split and domain-adversarial temporal convolutional network. The former segregates training data into temporal domains characterized by the most significant distribution differences, while the latter aims is designed to learn shared knowledge from the segmented temporal domains. The shared knowledge is independent of distribution shifts, which can be generalized to future testing well. Specifically, the domain-adversarial temporal convolutional network is composed of a base model incorporating temporal convolutional network and a domain classifier. Both components are jointly optimized by minimizing power prediction loss and maximizing temporal domain classification loss. The data from actual wind turbines are employed to validate the proposed approach. Experimental results show that the superior performance of the proposed approach compared to benchmark models, achieving a remarkable 24.56% improvement in accuracy by introducing adversarial training.

A Photovoltaic-Powered Modified Multiport Converter for an EV Charger with Bidirectional and Grid Connected Capability Assist PV2V, G2V, and V2G

To reduce the burden of electric vehicle (EV) charging power requirements, photovoltaic (PV) infrastructure EV charging has grown in recent years. The Z-Source Inverter (ZSI) allows tapping the boosted DC and AC by adjusting the switching shoot-through. However, it has only one DC tapping, thus limiting multiport charging options. This can be overcome by splitting the boosting capacitors used at the load terminal, which supports multiple charging ports, enabling simultaneous charging of multiple EVs, thereby increasing capacity and improving overall system efficiency. This paper presents a novel PV-tied Adaptable Z-Source Inverter (AZSI) for multiport EV charging. The modified split capacitor Z-source impedance networks ensure power availability at the charging station by regulating PV generation and grid supply. The performance of the AZSI was evaluated with experimentations that achieved an efficiency of 93.8% with three charging ports. This work contributes to developing sustainable and efficient charging infrastructure to meet the growing demands of the electric vehicle market.

Features Split and Aggregation Network for Camouflaged Object Detection.

Higher standards have been proposed for detection systems since camouflaged objects are not distinct enough, making it possible to ignore the difference between their background and foreground. In this paper, we present a new framework for Camouflaged Object Detection (COD) named FSANet, which consists mainly of three operations: spatial detail mining (SDM), cross-scale feature combination (CFC), and hierarchical feature aggregation decoder (HFAD). The framework simulates the three-stage detection process of the human visual mechanism when observing a camouflaged scene. Specifically, we have extracted five feature layers using the backbone and divided them into two parts with the second layer as the boundary. The SDM module simulates the human cursory inspection of the camouflaged objects to gather spatial details (such as edge, texture, etc.) and fuses the features to create a cursory impression. The CFC module is used to observe high-level features from various viewing angles and extracts the same features by thoroughly filtering features of various levels. We also design side-join multiplication in the CFC module to avoid detail distortion and use feature element-wise multiplication to filter out noise. Finally, we construct an HFAD module to deeply mine effective features from these two stages, direct the fusion of low-level features using high-level semantic knowledge, and improve the camouflage map using hierarchical cascade technology. Compared to the nineteen deep-learning-based methods in terms of seven widely used metrics, our proposed framework has clear advantages on four public COD datasets, demonstrating the effectiveness and superiority of our model.

ALSS-YOLO: An Adaptive Lightweight Channel Split and Shuffling Network for TIR Wildlife Detection in UAV Imagery

ALSS-YOLO: An Adaptive Lightweight Channel Split and Shuffling Network for TIR Wildlife Detection in UAV Imagery

CSST-Net: Channel Split Spatiotemporal Network for Human Action Recognition

Temporal reasoning is crucial for action recognition tasks. The previous works use 3D CNNs to jointly capture spatiotemporal information, but it causes a lot of computational costs as well. To improve the above problems, we propose a general channel split spatiotemporal network (CSST-Net) to achieve effective spatiotemporal feature representation learning. The CSST module consists of the grouped spatiotemporal modeling (GSTM) module and the parameter-free feature fusion (PFFF) module. The GSTM module decomposes features into spatial and temporal parts along the channel dimension in parallel, which focuses on spatial and temporal clues respectively. Meanwhile, we utilize the combination of group-wise convolution and point-wise convolution to reduce the number of parameters in the temporal branch, thus alleviating the overfitting of 3D CNNs. Furthermore, for the problem of spatiotemporal feature fusion, the PFFF module performs the recalibration and fusion of spatial and temporal features by a soft attention mechanism, without introducing extra parameters, thus ensuring the correct network information flow effectively. Finally, extensive experiments on three benchmark databases (Sth-Sth V1, Sth-Sth V2, and Jester) indicate that the proposed CSST-Net can achieve competitive performance compared to existing methods, and significantly reduces the number of parameters and FLOPs of 3D CNNs baseline.

A Novel ST-YOLO Network for Steel-Surface-Defect Detection.

Recent progress has been made in defect detection using methods based on deep learning, but there are still formidable obstacles. Defect images have rich semantic levels and diverse morphological features, and the model is dynamically changing due to ongoing learning. In response to these issues, this article proposes a shunt feature fusion model (ST-YOLO) for steel-defect detection, which uses a split feature network structure and a self-correcting transmission allocation method for training. The network structure is designed to specialize the process of classification and localization tasks for different computing needs. By using the self-correction criteria of adaptive sampling and dynamic label allocation, more sufficiently high-quality samples are utilized to adjust data distribution and optimize the training process. Our model achieved better performance on the NEU-DET datasets and the GC10-DET datasets and was validated to exhibit excellent performance.

Research on Digital Management and Collaboration Platform for Civil Aircraft Manufacturing Logistics Scenarios

Abstract In this paper, the overall architecture design of the platform is proposed, and the data architecture and business architecture of the platform are discussed. On this basis, the logistics path planning method of digital collaboration is proposed, including the aspects of path demand splitting and distribution vehicle spatio-temporal network, and the demand splittable logistics path planning model is considered. Finally, the simulation of the workshop logistics dynamic scheduling example is designed, and the dynamic scheduling optimization results are analyzed and verified. The results show that: through optimization, its time penalty cost f 1 = 801.38, the total distance traveled by the delivery trolley f 2 = 1709m, and the total distance of empty load is 692.15 m, and the dispatching scheme increases the distance of empty load by 59 m . The research results of this paper can provide a reference for digital management and collaboration platform for civil aircraft manufacturing enterprises and provide a reference for research in related fields.

SplitPlace: AI Augmented Splitting and Placement of Large-Scale Neural Networks in Mobile Edge Environments

In recent years, deep learning models have become ubiquitous in industry and academia alike. Deep neural networks can solve some of the most complex pattern-recognition problems today, but come with the price of massive compute and memory requirements. This makes the problem of deploying such large-scale neural networks challenging in resource-constrained mobile edge computing platforms. To solve this, a promising solution is to split resource-hungry neural networks into lightweight disjoint smaller components for pipelined distributed processing. At present, there are two main approaches to do this: semantic and layer-wise splitting. The former partitions a neural network into parallel disjoint models that produce a part of the result, whereas the latter partitions into sequential models that produce intermediate results. However, there is no intelligent algorithm that decides which splitting strategy to use and places such modular splits to edge nodes for optimal performance. To combat this, this work proposes a novel AI-driven online policy, SplitPlace, that uses Multi-Armed-Bandits to intelligently decide between layer and semantic splitting strategies based on the input task's service deadline demands. SplitPlace places such neural network split fragments on mobile edge devices using decision-aware reinforcement learning for efficient and scalable computing. Moreover, SplitPlace fine-tunes its placement engine to adapt to volatile environments. Our experiments on physical mobile-edge environments with real-world workloads show that SplitPlace can significantly improve the state-of-the-art in terms of average response time, deadline violation rate, inference accuracy, and total reward by up to 46, 69, 3 and 12 percent respectively.

Sparse Bandit Learning Based Location Management for Space-Ground Integrated Networks

The Space-Ground Integrated Network (SGIN) concept constitutes a promising solution for providing seamless global coverage. However, the mobility of satellites and wireless terminals imposes unprecedented challenges on the location management in SGIN. We tackle this challenge by conceiving a split identifier (ID) and Network Address (NA) based design for providing natural mobility support, and characterize the ID-NA mapping allocation problem by exploiting the storage capacity of both Geostationary Earth Orbit Satellites (GEOSs) and Low Earth Orbiting Satellites (LEOSs) to form a spatially distributed binding resolution system and optimize the caching reward in each LEOS. By considering the large quantity of ID-NA mapping and the sparsity of popular mapping having positive mean caching rewards, we formulate the mapping allocation problem as a sparse Multi-Armed Bandit (MAB) learning procedure, where the mappings are treated as the arms and the LEOSs act as the players. A distributed learning algorithm, namely the Sparse Upper confidence bound based Learning aided Caching algorithm (SULC), is proposed for estimating the mean caching rewards of mappings and selecting the optimal mappings for caching. Moreover, we derive a sub-linear upper bound of the cumulative learning regret to prove the learning efficiency of the proposed SULC. Extensive simulations have been conducted to show that the proposed SULC can quickly identify the popular mappings and provide near-optimal content hit rates. In contrast with the existing solutions, SULC has higher caching rewards and can significantly reduce the cumulative regret after a short period of learning.

COVID-19 Diagnosis and Detection Based on Deep Learning Models

The COVID-19 pandemic has caused widespread illness and death since its emergence in 2019. This study examines various Artificial Intelligence (AI) techniques for diagnosing and predicting COVID-19. One such method is the Hust 19 model, which employs a hybrid learning architecture of CNN and DNN models. The model divides CT scans into three types that identify COVID-19-related imaging features and implements a deep learning framework based on the VGG16 architecture. Another group of researchers developed a deep learning-based COVID-19 diagnostic system using multi-class and multi-center data, segmenting lungs and identifying COVID-19 infection slices. They evaluated the model's accuracy using Receiver Operating Characteristic (ROC) curves. A third group developed a deep neural network based on DenseNet121, standardizing input CXR images through anatomical landmark detection and registration, and segmenting lung lesions to diagnose pneumonia. A final group developed a three-dimensional deep learning model called COVNet, which takes CT images as input, extracts features from each slice using the ResNet50 backbone, merges the maximum features obtained by the AI model, and generates classification predictions for the entire CT scan. They also proposed a multi-decoder split network to improve the model's accuracy and efficiency. Experimental results show that the Deep learning AI system model and COVNet model are relatively good, with average sensitivity and specificity. The remaining models, particularly Hust-19, show prominent specificity, but high specificity leads to low sensitivity, making the overall model imbalanced. These AI diagnostic models are just the beginning, and there may be more inventions and creations in the future.

Residual Deformable Split Channel and Spatial U-Net for Automated Liver and Liver Tumour Segmentation.

Accurate segmentation of the liver and liver tumour (LT) is challenging due to its hazy boundaries and large shape variability. Although using U-Net for liver and LT segmentation achieves better results than manual segmentation, it loses spatial and channel features during segmentation, leading to inaccurate liver and LT segmentation. A residual deformable split depth-wise separable U-Net (RDSDSU-Net) is proposed to increase the accuracy of liver and LT segmentation. The residual deformable convolution layer (DCL) with deformable pooling (DP) is used in the encoder as an attention mechanism to adaptively extract liver and LT shape and position characteristics. Afterward, a convolutional spatial and channel features split graph network (CSCFSG-Net) is introduced in the middle processing layer to improve the expression capability of the liver and LT features by capturing spatial and channel features separately and to extract global contextual liver and LT information from spatial and channel features. Sub-pixel convolutions (SPC) are used in the decoder section to prevent the segmentation results from having a chequerboard artefact effect. Also, the residual deformable encoder features are combined with the decoder through summation to avoid increasing the number of feature maps (FM). Finally, the efficiency of the RDSDSU-Net is evaluated on the 3DIRCADb and LiTS datasets. The DICE score of the proposed RDSDSU-Net achieved 98.21% for liver segmentation and 93.25% for LT segmentation on 3DIRCADb. The experimental outcomes illustrate that the proposed RDSDSU-Net model achieved better segmentation results than the existing techniques.

Edge-Enhanced QoS Aware Compression Learning for Sustainable Data Stream Analytics

Existing Cloud systems involve large volumes of data streams being sent to a centralised data centre for monitoring, storage and analytics. However, migrating all the data to the cloud is often not feasible due to cost, privacy, and performance concerns. However, Machine Learning (ML) algorithms typically require significant computational resources, hence cannot be directly deployed on resource-constrained edge devices for learning and analytics. Edge-enhanced compressive offloading becomes a sustainable solution that allows data to be compressed at the edge and offloaded to the cloud for further analysis, reducing bandwidth consumption and communication latency. The design and implementation of a learning method for discovering compression techniques that offer the best QoS for an application is described. The approach uses a novel modularisation approach that maps features to models and classifies them for a range of Quality of Service (QoS) features. An automated QoS-aware orchestrator has been designed to select the best autoencoder model in real-time for compressive offloading in edge-enhanced clouds based on changing QoS requirements. The orchestrator has been designed to have diagnostic capabilities to search appropriate parameters that give the best compression. A key novelty of this work is harnessing the capabilities of autoencoders for edge-enhanced compressive offloading based on portable encodings, latent space splitting and fine-tuning network weights. Considering how the combination of features lead to different QoS models, the system is capable of processing a large number of user requests in a given time. The proposed hyperparameter search strategy (over the neural architectural space) reduces the computational cost of search through the entire space by up to 89%. When deployed on an edge-enhanced cloud using an Azure IoT testbed, the approach saves up to 70% data transfer costs and takes 32% less time for job completion. It eliminates the additional computational cost of decompression, thereby reducing the processing cost by up to 30%.

A semi-analytical solution to the stress intensity factors of branched cracks

Crack branching is ubiquitous in engineering practice, and it often takes place when a crack is subjected to dynamic stress fields, or runs into heterogeneous regions. The mechanical analysis of branched cracks is of great significance in safety analysis and crack-path engineering. In this work we developed a theoretical method to calculate the stress intensity factors (SIFs) of branched cracks. By employing both Schwarz–Christoffel mapping and Muskhelishvili approach, we present an asymptotic approximation for the conformal mapping and SIFs of arbitrary branched cracks are then readily derived. We further demonstrate the convenience of this analytical approach to obtain the SIFs of forked crack as well as four-branched cracks. The theoretical solutions are validated by using finite-element simulations. It is shown that the semi-analytical approach agrees well with the FEM calculations on SIFs. The analytical methods supply a general way to solve the SIFs and therefore the energy release rate of branched cracks. It can then be adopted to understand crack splitting and crack network engineering.

Whole genome resequencing reveal patterns of genetic variation within Colletotrichum acutatum species complex from rubber trees in China

The Colletotrichum acutatum species complex possesses a diverse number of important traits, such as a wide host range and host preference, different modes of reproduction, and different strategies of host infection. Research using comparative genomics has attempted to find correlations between these traits. Here, we used multi-locus techniques and gene genealogical concordance analysis to investigate the phylogenetic relationships and taxonomic status of the Colletotrichum acutatum species complex using field isolates obtained from rubber trees. The results revealed that the dominant species was C. australisinense, followed by C. bannaense, while strain YNJH17109 was identified as C. laticiphilum. The taxonomic status of strains YNLC510 and YNLC511 was undetermined. Using whole-genome single nucleotide polymorphism data to analyze population structure, 18 strains of C. australisinense were subsequently divided into four populations, one of which was derived from an admixture of two populations. In addition, the strains LD1687, GD1628, and YNLC516, did not belong to any populations, and were considered to be admixtures of two or more populations. A split decomposition network analysis also provided evidence for genetic recombination within Colletotrichum acutatum species complex from rubber trees in China. Overall, a weak phylogeographic sub-structure was observed. Analysis also revealed significant differences in morphological characters and levels of virulence between populations.

Split frequency attention network for single image deraining

Split frequency attention network for single image deraining