Backbone Network Research Articles

Using Artificial Neural Networks to Evaluate the Capacity and Cost of Multi-Fiber Optical Backbone Networks

A possible solution to address the enormous increase in traffic demands faced by network operators is to rely on multi-fiber optical backbone networks. These networks use multiple optical fibers between adjacent nodes, and, when properly designed, they are capable of handling petabits of data per second (Pbit/s). In this paper, an artificial neural network (ANN) model is investigated to estimate both the capacity and cost of a multi-fiber optical network. Furthermore, a fiber assignment algorithm is also proposed to complement the network design, enabling the generation of datasets for training and testing of the developed ANN model. The model consists of three layers, including one hidden layer with 50 hidden units. The results show that for a large network, such as one with 100 nodes, the model can estimate performance metrics with an average relative error of less than 0.4% for capacity and 4% for cost, while achieving a computation time nearly 800 times faster than the heuristic approach used in network simulation. Additionally, the network capacity is around 5 Pbit/s.

Multi-architecture optimization of pipeline inner wall defect detection algorithm based on YOLOv8

As traditional computer vision technology struggles to meet the demands for accurate detection of modern supply pipes. A detection model based on an improved YOLOv8 network has been proposed. First, the Inverted Residual Mobile Block (iRMB) is integrated into the backbone network. This effectively enhances the feature representation capability for extracting image defects. Next, a shift-wise operator is introduced to simulate the effects of using a large convolutional kernel at a lower computational cost while improving performance. Finally, GSConv replaces the Conv layer in the neck network, and the VoVGSCSP module substitutes the C2f module in the neck network to further enhance the algorithm’s detection accuracy for corroded regions. Experimental results demonstrate that the mean Average Precision (mAP) of the improved algorithm reaches 95.0 % on the dataset of pipeline inner wall corrosion defects. This provides an accurate method for the intelligent identification of corrosion defects on the pipeline dataset.

Research on Irrigation Grade Discrimination Method Based on Semantic Segmentation

As one of China’s major grain crops, wheat has a high demand for water resources, making it susceptible to drought stress. Traditional irrigation evaluation methods are often based on experience and rule-based calculations, which struggle to cope with complex environmental factors and dynamic changes in crop needs. With technological advancements, deep learning-based research methods, characterized by their strong data-driven analytical capabilities, are expected to improve the accuracy of evaluation results. This paper focuses on the irrigation demand assessment of winter wheat farmland, aiming to explore a new regional-scale irrigation demand assessment method based on deep learning. By establishing samples of different irrigation evaluation levels, this study seeks to better meet the requirements of irrigation demand assessment. For the problem of regional-scale irrigation-level discrimination, the Convolutional Network Attention(CONAT) module was proposed to optimize the backbone network structure of the Mask2Former model. To tackle issues related to data imbalance and underfitting across certain categories, a loss function tailored for imbalanced sample distributions was implemented, accompanied by enhancements to the training scheme. By contrasting this refined model with alternative methods for discriminating irrigation levels, the viability of this approach was showcased.

A Doodle-Based Control for Characters in Story Visualization

We propose a story visualization technique that allows users to control the arrangement, poses, and styles of characters in a scene based on user-input doodle sketches. Our method utilizes a text encoder to process scene prompts and an image encoder to handle doodle sketches, generating inputs for a predefined scene generation model. Furthermore, we achieve efficient model training by fine-tuning the backbone network by applying a small dataset and employing a LoRA-based fine-tuning technique. We demonstrate that our method can generate characters with various poses and styles from doodle sketches, and it can validate the advantages of our approach by comparing it with the results from other story visualization studies.

Diabetic retinopathy grading based on multi-scale residual network and cross-attention module

Diabetic retinopathy (DR) is a severe effect of diabetes mellitus that mainly impacts the retinal tissue and carries a high risk of blindness. Ophthalmologists face challenges in assessing the severity of DR due to its complexity and time constraints. Consequently, there is an urgent need for the development of automated methods that employ retinal fundus images to detect DR. This study introduces a novel deep learning architecture that utilizes a multi-scale residual attention block (MSRAB) and a cross-attention block (CrAB) for DR grading. The proposed MSRAB employs a convolutional neural network (CNN) with diverse dilation rates to expand its field of view. MSRAB adaptively concentrate on pertinent features and incorporates a residual attention network to improve grading performance. Integrating the residual attention network enables MSRAB to prioritize critical characteristics in retinal fundus images and enhance performance. Similarly, the CrAB integrates channel and spatial attention mechanisms to capture inter-channel interactions and spatial dependencies within the input features. This comprehensive methodology enables the model to concentrate more efficiently on critical regions to discriminate irrelevant features and to capture interrelations across different channels and spatial regions. The proposed model employs a pre-trained backbone network to extract local and global features to capture complex features for precisely identifying DR. This process enhances model efficiency, which is helpful in methods with limited data and processing limitations. The suggested methodology surpasses the existing methodologies in accuracy, recall, precision, and Area Under the Curve.

Web-Based GIS Housing Marketing Information System in West Papua

Web GIS- Based Marketing Information System in Housing Sector of West Papua: A Case Study The methods of research are interviews, direct observation and documents sales report, regional maps). They note that the system offers a more direct and efficient means of accessing information from potential buyers, ultimately boosting marketing efficiency by as much as 45%. Moreover, it delivered a 30% higher sales conversion rate than traditional marketing channels. A combination of useful features such as interactive maps and location search functions helps potential buyers browse properties more easily, increasing the time they spend interacting with the platform and enhancing transaction chances. Many housing developers overcame challenges like the lack of internet backbone and a requirement for staff training to employ specialised skillsets. To needs, this research shows as an imperative using of Web gip to smooth the marketing and support strategic plan in housing construction. A recommendation for improving technological infrastructure and continuous training to ensure an effective utilization of the system is suggested. The results offer useful advice to housing developers looking to build innovative technology into their marketing strategy, thereby increasing efficiency and effectiveness within the housing sector.

RCT-YOLOv8: A Tuna Detection Model for Distant-Water Fisheries Based on Improved YOLOv8

With the development of distant-water fisheries, ship fishing and fish catch detection are now vital to modern fishing. Existing manual detection methods are prone to issues such as missed detections and false detections. Deep learning has enabled the deployment of detection models on shipboard devices, offering a new solution. However, many existing models are hindered by large parameters and computational complexity, making them unsuitable for shipboard use due to limited resources and costs onboard ships. To address these challenges, we propose the RCT-YOLOv8 model for tuna catch detection in this paper. Specifically, we adopt YOLOv8 as the base model and replace the network backbone with RepVGG network, which employs re-parameterized convolutions to enhance detection accuracy. Additionally, we incorporate coordinate attention at the end of the backbone to better aggregate channel-wise information. In the neck part, we introduce the contextual transformer (CoT) attention and propose the C2F-CoT model, which combines convolutional neural network with Transformer to capture global features, thereby improving detection accuracy and the effectiveness of feature propagation. We test multiple loss functions and select efficient intersection over union, which is more suitable for our algorithm. Furthermore, to adapt to devices with limited computational resources, we utilize the dependency-graph-based pruning method to compress the network model. Compared to the base network, the pruned model achieves a 9.8% increase in detection accuracy while reducing parameters and computational complexity by 40% and 35.8%, respectively. Compared to various algorithms, the pruned model demonstrates the highest detection accuracy, lowest parameter count, and lowest computational complexity, achieving optimal results at all fronts.

Multi-Scenario Remote Sensing Image Forgery Detection Based on Transformer and Model Fusion

Recently, remote sensing image forgery detection has received widespread attention. To improve the detection accuracy, we build a novel scheme based on Transformer and model fusion. Specifically, we model this task as a binary classification task that focuses on global information. First, we explore the performance of various excellent feature extraction networks in this task under the constructed unified classification framework. On this basis, we select three high-performance Transformer-based networks that focus on global information, namely, Swin Transformer V1, Swin Transformer V2, and Twins, as the backbone networks and fuse them. Secondly, considering the small number of samples, we use the public ImageNet-1K dataset to pre-train the network to learn more stable feature expressions. At the same time, a circular data divide strategy is proposed, which can fully utilize all the samples to improve the accuracy in the competition. Finally, to promote network optimization, on the one hand, we explore multiple loss functions and select label smooth loss, which can reduce the model’s excessive dependence on training data. On the other hand, we construct a combined learning rate optimization strategy that first uses step degeneration and then cosine annealing, which reduces the risk of the network falling into local optima. Extensive experiments show that the proposed scheme has excellent performance. This scheme won seventh place in the “Forgery Detection in Multi-scenario Remote Sensing Images of Typical Objects” track of the 2024 ISPRS TC I contest on Intelligent Interpretation for Multi-modal Remote Sensing Application.

DV-DETR: Improved UAV Aerial Small Target Detection Algorithm Based on RT-DETR.

For drone-based detection tasks, accurately identifying small-scale targets like people, bicycles, and pedestrians remains a key challenge. In this paper, we propose DV-DETR, an improved detection model based on the Real-Time Detection Transformer (RT-DETR), specifically optimized for small target detection in high-density scenes. To achieve this, we introduce three main enhancements: (1) ResNet18 as the backbone network to improve feature extraction and reduce model complexity; (2) the integration of recalibration attention units and deformable attention mechanisms in the neck network to enhance multi-scale feature fusion and improve localization accuracy; and (3) the use of the Focaler-IoU loss function to better handle the imbalanced distribution of target scales and focus on challenging samples. Experimental results on the VisDrone2019 dataset show that DV-DETR achieves an mAP@0.5 of 50.1%, a 1.7% improvement over the baseline model, while increasing detection speed from 75 FPS to 90 FPS, meeting real-time processing requirements. These improvements not only enhance the model's accuracy and efficiency but also provide practical significance in complex, high-density urban environments, supporting real-world applications in UAV-based surveillance and monitoring tasks.

Recognition of Maize Tassels Based on Improved YOLOv8 and Unmanned Aerial Vehicles RGB Images

The tasseling stage of maize, as a critical period of maize cultivation, is essential for predicting maize yield and understanding the normal condition of maize growth. However, the branches overlap each other during the growth of maize seedlings and cannot be used as an identifying feature. However, during the tasseling stage, its apical ear blooms and has distinctive features that can be used as an identifying feature. However, the sizes of the maize tassels are small, the background is complex, and the existing network has obvious recognition errors. Therefore, in this paper, unmanned aerial vehicle (UAV) RGB images and an improved YOLOv8 target detection network are used to enhance the recognition accuracy of maize tassels. In the new network, a microscale target detection head is added to increase the ability to perceive small-sized maize tassels; In addition, Spatial Pyramid Pooling—Fast (SPPF) is replaced by the Spatial Pyramid Pooling with Efficient Layer Aggregation Network (SPPELAN) in the backbone network part to connect different levels of detailed features and semantic information. Moreover, a dual-attention module synthesized by GAM-CBAM is added to the neck part to reduce the loss of features of maize tassels, thus improving the network’s detection ability. We also labeled the new maize tassels dataset in VOC format as the training and validation of the network model. In the final model testing results, the new network model’s precision reached 93.6% and recall reached 92.5%, which was an improvement of 2.8–12.6 percentage points and 3.6–15.2 percentage points compared to the mAP50 and F1-score values of other models. From the experimental results, it is shown that the improved YOLOv8 network, with high performance and robustness in small-sized maize tassel recognition, can accurately recognize maize tassels in UAV images, which provides technical support for automated counting, accurate cultivation, and large-scale intelligent cultivation of maize seedlings.

Research on litchi image detection in orchard using UAV based on improved YOLOv5

The growth environment of litchis is complex, and there is a tendency for occlusion between litchis, which makes litchi detection difficult. The use of UAVs to detect litchi in complex natural environments is important for litchi yield estimation. In this paper, an improved You Only Look Once version 5 (YOLOv5) model YOLO-Litchi based on litchi images captured by UAVs is proposed, which can be used for litchi detection and yield estimation in orchards. We establish a dataset of litchi in the litchi orchard of South China Agricultural University (SCAU) using a DJI Elf 4 generation drone, and chose the YOLOv5s model as the base version of the improved model for litchi detection in complex natural environments. First, we add a small target detection layer for the characteristics of litchi targets in UAV images, so that the model can localize smaller targets and improve the detection performance of small targets. Second, we replace the C3 module C3TR with the C3 module C3TR incorporating a transformer at the end of the network backbone and add the ECA attention (Efficient Channel Attention) mechanism after the C3TR module to reduce the interference of invalid information in the background and enhance the ability of the network to extract litchi feature information. Finally, we replace the SiLU activation function in the original model with MemoryEfficientMish, a more efficient version of the Mish activation function that customizes the forward propagation and backpropagation, making its backpropagation more efficient. The effectiveness of the improvement is verified experimentally. The results show that the mAP and recall of the proposed model in this paper are improved by 1.5 % and 2.4 %, respectively, compared with the unimproved YOLOv5 network. Compared with object detection models such as SSD and YOLOv7, the model proposed in this paper also has better performance. In addition, our proposed model maintains the FPS of the model at 45.87 and the model size at 20.7 MB while improving the detection effect, which still meets the requirements of real-time detection as well as model lightweighting. In summary, the proposed model successfully balances speed, model size, and accuracy, meeting the needs of agricultural litchi detection. The improved litchi detection model is more robust, has higher detection accuracy, and is suitable for litchi recognition and yield estimation in field environments, which helps to refine the management of litchi plantations.

Communications, media and Internet Concentration in the South Cone 2019-2021

The article analyzes market concentration in communication, internet and media industries in Argentina, Brazil and Chile between 2019- 2021 using CR4 and HHI indexes. It concludes that the Network Media Economy is highly concentrated. The Internet Core applications sectors operate in an almost monopolistic position. The telecommunications and Internet infrastructure sector collects the majority of revenues of the Network Media Economy with very high concentration in Wireless, Wireline and a little less, but still very high, in Multichannel video distribution and ISPs. Media sectors show significant concentration in audiovisual services and low concentration in online press.

Assessing the impact of structural modifications in the construction of surveillance network for Peste des petits ruminants disease in Nigeria: The role of backbone and sentinel nodes.

Peste des petits ruminants (PPR) is a highly contagious disease affecting mainly sheep and goats. Livestock movements contribute to the spread of the disease by introducing it to naive areas or exposing susceptible animals to it in infected regions. Because of its socio-economic impact, the Food and Agriculture Organisation (FAO) and the World Organization for Animal Health (WOAH) have set the goal to eradicate it by 2030, one of the key steps being the improvement of surveillance networks. The present study aimed to provide tools to identify areas that could serve as sentinel nodes, i.e. areas that may be rapidly infected at the onset of epidemics. Using data from a market survey conducted in Northern Nigeria, we reconstructed the small ruminants mobility network and simulated the diffusion of PPR virus through animal movement. From the analysis of simulation outcomes, we investigated which nodes could act as sentinel nodes under specific conditions for disease transmission. We considered several modified networks to get around the problem of data only being available for part of the overall network structure and to account for potential errors made during the field study. For each configuration, we simulated the spread of PPR using a stochastic Susceptible-Infectious (SI) model based on animal movements to assess the epidemics' extent and the presence of recurrent patterns to identify potential sentinel nodes. We extracted the backbone of the reference network and checked for the presence of sentinel nodes within it. We investigated how the origin (seed) of the epidemics could affect the propagation pattern by comparing and grouping seeds based on their respective transmission paths. Results showed that the isolated backbone contains 45% of sentinel nodes that remain stable or undergo only minor changes in 9 out of 11 configurations. On top of that, the characteristics of sentinel nodes identified in the backbone are not influenced by the severity of the disease. The H index, in-degree, and eigenvector are the most essential variables. This study provides an overview of the major axes of animal movements in Nigeria and the most vulnerable locations that should be prioritized for monitoring livestock diseases like PPR.

Caricature-visual face recognition based on jigsaw solving and modal decoupling

In recent years, face recognition technology has made significant progress in the field of real visual images, yet face recognition involving caricature-visual images remains a challenge due to the exaggerated and unrealistic features of caricature faces. To tackle this issue, this paper introduces the Caricature-visual Face Recognition Model Based on Jigsaw Solving and Modal Decoupling (CVF-JSM). The CVF-JSM consists of two modules: feature extraction and decoupling. The feature extraction module incorporates a graph attention network at the intermediate stage of the backbone network, which constructs and solves jigsaw puzzles to enable the network to extract shape features. The feature decoupling module features a three-branch structure that divides the features into modal and identity features. The real and caricature face recognition branches separate identity features for recognition through parameter sharing and orthogonality constraints. The feature common subspace alignment branch maps the anchor image, as well as the positive and negative sample images, into a common subspace to isolate identity features. Subsequently, by aligning the features, it further refines the effective identity features. The experimental results conducted on multiple datasets demonstrate that the CVF-JSM model outperforms existing technologies in the realm of caricature-visual face recognition.

Fabric defect detection method based on multi-scale fusion attention mechanisms

Abstract Fabric defect detection is extremely important for the development of the textile industry, but the existing traditional image processing algorithms are not good enough to detect fabric defects, and the detection efficiency and accuracy of the classical deep learning model is not satisfactory, so this paper proposes an improved fabric defect detection method based on multi-scale fusion of attention mechanism YOLOv7-PCBS. Based on the YOLOv7 network structure, some of the standard convolutions of the backbone network are replaced with Partial Convolution modules, which reduces the amount of network computation and improves the network detection speed; add coordinate attention to enhance the ability of extracting the positional features of tiny defects in fabrics; reconfiguration of the SPPCSPC module to improve small target detection; optimization of Bidirectional Feature Pyramid Network (BiFPN) and design of Tiny- BiFPN for simple and fast multi-scale feature fusion; finally, a novel loss function SIoU with angular loss is introduced to facilitate the fitting of the true and predicted frames and enhance the accuracy of defect prediction. The results show that the algorithm achieves a mAP value of 94.4% on the detection of defects in solid-colored fabrics of six denim materials, which is an improvement of 15.1% compared to the original YOLOv7 algorithm, while the model achieves a frame rate of 59.5 per second. Compared with other traditional deep learning algorithms SSD and Faster-RCNN, the detection accuracies are improved by 21.6% and 15.2%, and the FPS values are improved by 78.1% and 101.0%, respectively. Therefore, the YOLOv7-PCBS fabric defect detection algorithm proposed in this paper makes the fabric defect detection results more accurate while realizing lightweight, which provides an important technical reference for the subsequent improvement of textile quality.

Based on Improved Lightweight YOLOv8 for Vehicle Detection

In urban road traffic, vehicle detection in intelligent transportation system can effectively improve road traffic operation efficiency and road safety. Based on this goal, this paper is based on the YOLOv8 framework, using the improved lightweight YOLOv8 model to realize vehicle detection, the method replaces the Backbone network in the YOLOv8n model with a more efficient and lightweight EfficientViT network, and further adds the CBAM attention mechanism in Neck to enhance the detection precision, and later, Conv is replaced with the GhostConv model in Head to reduce the number of parameters and increase the speed of detection. This model greatly improves the detection precision and efficiency. We validate our approach on the UA-DETRAC vehicle detection dataset, and the experimental results show that the proposed detection model Precision reaches 91.17%, mAP@0.5 reaches 75.45%, and Recall reaches about 70.01%. Compared with the original YOLOv8n model, Precision improves by 4.6%, mAP@0.5 improves by about 4.3%, and Recall improves by about 9.1%. Through experimental validation, the model performs well in detecting various complex road traffic scenarios, and this innovative approach can provide strong support for the development of intelligent transportation systems, such as deployed on devices with limited mobile hardware resources, and is expected to make further breakthroughs in future applications.

Research on coal gangue recognition method based on SFD-YOLOv5s

ABSTRACT Aiming at the current coal gangue recognition model’s difficulty in recognizing aggregated coal gangue targets, which leads to the problems of misdetections, omissions, and inaccurate localization, we propose a coal gangue recognition method based on the SFD-YOLOv5s model. The SE attention mechanism is improved in incorporating maximum pooling and introduced into the backbone network of the YOLOv5s model to improve the model’s ability to extract the critical feature information of the coal gangue. Furthermore, to improve the recognition ability of the YOLOv5s model for aggregated targets, we introduce the Focal-EIoU loss function and the Dyhead. The experimental results show that the overall performance of the SFD-YOLOv5s model has surpassed other mainstream object detection models of the YOLO series. The SFD-YOLOv5s model has the highest precision, recall, and mean average precision. The precision reaches 97.9%, the recall is 96.2%, and the mean average precision is 94.5%, which is 8.7%, 8.9%, and 9.0% higher compared with the YOLOv5s model, respectively. The FPS is 96.15, which can fully meet the demand for real-time detection of coal gangue. Meanwhile, the SFD-YOLOv5s model has the best recognition effect of aggregated coal gangue targets.

Facile preparation of Co3O4/Super P/multi-walled carbon nanotube films for high-performance Li-ion battery anodes

This study prepared Co3O4/Super P/Multi-walled carbon nanotube (MWCNT) free-standing films through a simple vacuum filtration method. The free-standing films showed that Co3O4 particles were well dispersed within the Super P and MWCNT network backbone without deforming oxide materials, confirmed by scanning electron microscopy (SEM) images and X-ray diffraction (XRD) patterns. Employing Co3O4/Super P/MWCNT free-standing films as an anode electrode in lithium-ion batteries is superior to conventional electrode procedures because a binder is not required, and the current collector and laminate process are required. Moreover, the Co3O4/Super P/MWCNT electrodes showed remarkable capacity and rate capability compared to that of conventional laminated Co3O4/carbon composites. The enhanced electrochemical performance of Co3O4/Super P/MWCNT electrodes could be attributed to the extraordinary electrical conduction of MWCNT in nature and its stable connection between the conductive substrate and active materials. Furthermore, this connection offers the integrity of the electrode despite volume variation during Lithium (Li)-ion intercalation and deintercalation.

Interpretable rough neural network for lung nodule diagnosis

Computer-aided diagnosis (CAD) systems based on deep learning have shown significant potential in lung nodule diagnosis, providing substantial assistance to medical professionals. However, the inherent lack of interpretability in deep learning models and the uncertainty of annotations limit their widespread application. We propose that uncertain annotations actually imply additional valuable information that can enhance both model performance and interpretability. To address these challenges, we have developed a novel soft computing methodology integrating rough sets with deep neural networks. Firstly, this methodology employs rough sets to process uncertain region of interest (ROI) annotations into upper and lower approximations. Secondly, a novel rough neuron is designed to predict these approximations. Thirdly, the newly proposed region-constraint strategy embeds interpretable radiological domain knowledge into the neural network. Finally, this methodology proposes interpretation curves and regional consistency metrics to quantitatively evaluate the model’s interpretability. We conducted extensive comparison experiments on LIDC-IDRI and LNDb public benchmarks. Detailed experimental results demonstrate that by maximally retaining uncertain samples, the proposed method achieves classification accuracies of 84.6% and 89.74%, and mean absolute errors of 0.4988 and 0.5208 in attribute prediction, representing improvements of 3.4% and 2.5%, respectively, over the backbone networks.

GreenFruitDetector: Lightweight green fruit detector in orchard environment.

Detecting green fruits presents significant challenges due to their close resemblance in color to the leaves in an orchard environment. We designed GreenFruitDetector, a lightweight model based on an improved YOLO v8 architecture, specifically for green fruit detection. In the Backbone network, we replace ordinary convolution with Deformable Convolution to enhance the extraction of geometric features. Additionally, we designed MCAG-DC (Multi-path Coordinate Attention Guided Deformer Convolution) to replace the convolution in C2f, enhancing the Backbone's feature extraction capability when encountering occlusion problems. For the Neck part of the algorithm, we designed a Fusion-neck structure that integrates spatial detail information from feature maps at different scales, thereby enhancing the network's ability to extract multi-scale information. Additionally, we devised a new detection head that incorporates multi-scale information, significantly improving the detection of small and distant objects. Finally, we applied channel pruning techniques to reduce the model size, parameter count, and FLOPs to 50%, 55%, and 44% of the original, respectively. We trained and evaluated the improved model on three green fruit datasets. The accuracy of the improved model reached 94.5%, 84.4%, and 85.9% on the Korla Pear, Guava, and Green Apple datasets, respectively, representing improvements of 1.17%, 1.1%, and 1.77% over the baseline model. The mAP@0.5 increased by 0.72%, 6.5%, and 0.9%, respectively, and the recall rate increased by 1.97%, 1.1%, and 0.49%, respectively.