Backbone Network Research Articles

Improved lightweight YOLOv5 based on ShuffleNet and its application on traffic signs detection.

Traffic signs detection is an important and challenging task in intelligent driving perception system. This paper proposes an improved lightweight traffic signs detection framework based on YOLOv5. Firstly, the YOLOv5's backbone is replaced with ShuffleNet v2, which simplifies the calculation complexity and reduces the parameters of backbone network. Secondly, aiming at the problem of inconspicuous traffic sign characteristics in complex road environment, we use the CA attention mechanism in this paper to improve the saliency of the object. Finally, aiming at the large-scale difference between the traffic signs and the high proportion of small objects, we design the BCS-FPN to fuse multi-scale features and improve the representation ability of the small-scale objects. The TT-100K dataset is also analyzed and the dataset is collated. We test on the collated TT-100K dataset for the improved YOLOv5 in this paper. And the results show that compared with YOLOv5s, the mAP of our algorithm is equivalent to that of YOLOv5s, and the speed is improved by 20.8%. This paper also has carried on the experiment on embedded devices, experimental results show that our framework in computing power less embedded devices has a better effect.

Implementing Anchor Free Model for Social Distancing Detection on FPGA Board

Since 2019, the world has known a global pandemic caused by the COVID-19 virus. The epidemic was spreading very fast and many precautions should be respected to fight the disease. In order to limit the CVID-19 spread, different ways can be adopted. Social distancing is one of the most important cautions that should be respected. It means keeping a safe social distance between persons in order to avoid COVID-19 contagion. The social distance is about one meter at least. Building new tools used to perform social distance system present a very challenging task. We propose in this paper to build a social distancing system based on one-stage neural networks. The proposed system is developed based on an improved version FCOS model. In order to ensure an embedded implementation of the proposed work, we used EfficientNet v1 as a network backbone and we applied compression techniques to reduce the model complexity and computation resources. The inference stage of the model has been performed on a ZCU 102 board. Training and testing experiments have demonstrated the efficiency of the proposed work in terms of accuracy as well as processing time.

Correction: Freitas, A.; Pires, J. Investigating the Impact of Topology and Physical Impairments on the Capacity of an Optical Backbone Network. Photonics 2024, 11, 342

In the original publication [...]

Improve the Performance of SAR Ship Detectors by Small Object Detection Strategies

Although advanced deep learning techniques have significantly improved SAR ship detection, accurately detecting small ships remains challenging due to their limited size and the few appearance and geometric clues available. In order to solve this problem, we propose several small object detection strategies. The backbone network uses space-to-depth convolution to replace stride and pooling. It reduces information loss during down-sampling. The neck integrates multiple layers of features globally and injects global and local information into different levels. It avoids the inherent information loss of traditional feature pyramid networks and strengthens the information fusion ability without significantly increasing latency. The proposed intersection-of-union considers the center distance and scale of small ships specifically. It reduces the sensitivity of intersection-of-union to positional deviations of small ships, which is helpful for training toward small ships. During training, the smaller the localization loss of small ships, the greater their localization loss gains are. By this, the supervision of small ships is strengthened in the loss function, which can make the losses more biased toward small ships. A series of experiments are conducted on two commonly used datasets, SSDD and SAR-Ship-Dataset. The experimental results show that the proposed method can detect small ships successfully and thus improve the overall performance of detectors.

Improved Architecture and Training Strategies of YOLOv7 for Remote Sensing Image Object Detection

The technology for object detection in remote sensing images finds extensive applications in production and people’s lives, and improving the accuracy of image detection is a pressing need. With that goal, this paper proposes a range of improvements, rooted in the widely used YOLOv7 algorithm, after analyzing the requirements and difficulties in the detection of remote sensing images. Specifically, we strategically remove some standard convolution and pooling modules from the bottom of the network, adopting stride-free convolution to minimize the loss of information for small objects in the transmission. Simultaneously, we introduce a new, more efficient attention mechanism module for feature extraction, significantly enhancing the network’s semantic extraction capabilities. Furthermore, by adding multiple cross-layer connections in the network, we more effectively utilize the feature information of each layer in the backbone network, thereby enhancing the network’s overall feature extraction capability. During the training phase, we introduce an auxiliary network to intensify the training of the underlying network and adopt a new activation function and a more efficient loss function to ensure more effective gradient feedback, thereby elevating the network performance. In the experimental results, our improved network achieves impressive mAP scores of 91.2% and 80.8% on the DIOR and DOTA version 1.0 remote sensing datasets, respectively. These represent notable improvements of 4.5% and 7.0% over the original YOLOv7 network, significantly enhancing the efficiency of detecting small objects in particular.

Fabric Defect Detection Based on Improved Lightweight YOLOv8n

In response to the challenges posed by complex background textures and limited hardware resources in fabric defect detection, this study proposes a lightweight fabric defect detection algorithm based on an improved GSL-YOLOv8n model. Firstly, to reduce the parameter count and complexity of the YOLOv8n network, the GhostNet concept is used to construct the C2fGhost module, replacing the conventional convolution layers in the YOLOv8n structure with Ghost convolutions. Secondly, the SimAM parameter-free attention mechanism is embedded at the end of the backbone network to eliminate redundant background, enhance semantic information for small targets, and improve the network’s feature extraction capability. Lastly, a lightweight shared convolution detection head is designed, employing the scale layer to adjust features, ensuring the lightweight nature of the model while minimizing precision loss. Compared to the original YOLOv8n model, the improved GSL-YOLOv8n algorithm increases the mAP@0.5 by 0.60% to 98.29% and reduces model size, computational load, and parameter count by 66.7%, 58.0%, and 67.4%, respectively, meeting the application requirements for fabric defect detection in textile industry production.

Research on gear flank surface defect detection of automotive transmissions based on improved YOLOv8

Abstract In industrial production, the manufacturing processes may introduce defects on the gear flanks of transmission gears, potentially leading to premature failures and diminished performance. The early detection and precise assessment of surface defects on transmission gear flanks are critical for maintaining the safety, reliability, and cost-effectiveness of automobiles. At present, the principal approach for identifying defects on automotive transmission gear flanks predominantly involves manual visual inspections, supplemented by fluorescent magnetic particle testing. However, this approach suffers from low accuracy and efficiency. Consequently, this paper presents a defect detection algorithm that leverages an enhanced YOLOv8 model to facilitate the efficient detection of surface defects on automotive transmission gear flanks. Initially, the collected image data underwent data augmentation and exploratory analysis, which informed targeted enhancements. Subsequently, the YOLOv8 algorithm was thoroughly examined. The SPPE architecture was incorporated into the backbone network, and the DCNv4 module was integrated to boost the model's capability in detecting irregular defects. In the neck network, the BiFormer attention mechanism was implemented to enhance detection performance for small-scale defects. Moreover, the newly developed MASFF_Head structure was adopted as the detection head to augment detection efficacy for multi-scale defects. Additionally, the bounding box-loss function was substituted with the WIoU loss to improve performance on low-quality samples. Experimental results demonstrated that the mAP@0.5 of the refined YOLOv8 network model reached 86.1%, marking a 2.8% increase over the original model and significantly boosting detection accuracy. When compared to other deep learning models, the enhanced YOLOv8 model exhibits considerable superiority in terms of detection precision and efficiency. The P value and R value achieved were 82.9% and 80.8%, respectively, with a detection time of 21.6 milliseconds. This underscores the method's effectiveness and reliability in detecting automotive transmission gear defects, underscoring its pivotal role in facilitating automated detection processes on industrial production lines.

AL-MobileNet: a novel model for 2D gesture recognition in intelligent cockpit based on multi-modal data

As the degree of automotive intelligence increases, gesture recognition is gaining more attention in human-vehicle interaction. However, existing gesture recognition methods are computationally intensive and perform poorly in multi-modal sensor scenarios. This paper proposes a novel network structure, AL-MobileNet (MobileNet with Attention and Lightweight Modules), which can quickly and accurately estimate 2D gestures in RGB and infrared (IR) images. The innovations of this paper are as follows: Firstly, to enhance multi-modal data, we created a synthetic IR dataset based on real 2D gestures and employed a coarse-to-fine training approach. Secondly, to speed up the model's computation on edge devices, we introduced a new lightweight computational module called the Split Channel Attention Block (SCAB). Thirdly, to ensure the model maintains accuracy in large datasets, we incorporated auxiliary networks and Angle-Weighted Loss (AWL) into the backbone network. Experiments show that AL-MobileNet requires only 0.4 GFLOPs of computational power and 1.2 million parameters. This makes it 1.5 times faster than MobileNet and allows for quick execution on edge devices. AL-MobileNet achieved a running speed of up to 28 FPS on the Ambarella CV28. On both general datasets and our dataset, our algorithm achieved an average PCK0.2 score of 0.95. This indicates that the algorithm can quickly generate accurate 2D gestures. The demonstration of the algorithm can be reviewed in gesturebaolong.

Personnel Monitoring in Shipboard Surveillance Using Improved Multi-Object Detection and Tracking Algorithm.

Detecting and tracking personnel onboard is an important measure to prevent ships from being invaded by outsiders and ensure ship security. Ships are characterized by more cabins, numerous equipment, and dense personnel, so there are problems such as unpredictable personnel trajectories, frequent occlusions, and many small targets, which lead to the poor performance of existing multi-target-tracking algorithms on shipboard surveillance videos. This study conducts research in the context of onboard surveillance and proposes a multi-object detection and tracking algorithm for anti-intrusion on ships. First, this study designs the BR-YOLO network to provide high-quality object-detection results for the tracking algorithm. The shallow layers of its backbone network use the BiFormer module to capture dependencies between distant objects and reduce information loss. Second, the improved C2f module is used in the deep layer of BR-YOLO to introduce the RepGhost structure to achieve model lightweighting through reparameterization. Then, the Part OSNet network is proposed, which uses different pooling branches to focus on multi-scale features, including part-level features, thereby obtaining strong Re-ID feature representations and providing richer appearance information for personnel tracking. Finally, by integrating the appearance information for association matching, the tracking trajectory is generated in Tracking-By-Detection mode and validated on the self-constructed shipboard surveillance dataset. The experimental results show that the algorithm in this paper is effective in shipboard surveillance. Compared with the present mainstream algorithms, the MOTA, HOTP, and IDF1 are enhanced by about 10 percentage points, the MOTP is enhanced by about 7 percentage points, and IDs are also significantly reduced, which is of great practical significance for the prevention of intrusion by ship personnel.

Flying foxes optimization with reinforcement learning for vehicle detection in UAV imagery

Intelligent transportation systems (ITS) are globally installed in smart cities, which enable the next generation of ITS depending on the potential integration of autonomous and connected vehicles. Both technologies are being tested widely in various cities across the world. However, these two developing technologies are vital in allowing a fully automatic transportation system; it is necessary to automate other transportation and road components. Unmanned aerial vehicles (UAVs) or drones are utilized for many surveillance applications in the ITS. Detecting on-ground vehicles in drone images is significant for disaster rescue operations, traffic and parking management, and navigating uneven territories. This study presents a flying foxes optimization with deep learning-based vehicle detection and classification model on aerial images (FFODL-VDCAI) technique for ITS application. The main objective of the FFODL-VDCAI technique is to automate and accurately classify vehicles that exist in aerial images. Three primary processes are involved in the presented FFODL-VDCAI technique. Initially, the FFODL-VDCAI approach utilizes YOLO-GD (Ghost-Net and Depthwise convolution) for vehicle detection, where the YOLO-GD uses lightweight Ghost Net in place on the backbone network of YOLO-v4 and interchanges the conventional convolutional with depthwise separable convolutional and pointwise convolutional. Next, the FFO technique is used for hyperparameter tuning the Ghost Net technique. Finally, a deep Q-network (DQN) based reinforcement learning technique is used to classify detected vehicles effectively. A comprehensive simulation analysis of the FFODL-VDCAI methodology is conducted on the UAV image dataset. The performance validation of the FFODL-VDCAI methodology exhibited superior values of 96.15% and 92.03% under PSU and Stanford datasets concerning various aspects.

Brain tumor image segmentation method using hybrid attention module and improved mask RCNN

To meet the needs of automated medical analysis of brain tumor magnetic resonance imaging, this study introduces an enhanced instance segmentation method built upon mask region-based convolutional neural network. By incorporating squeeze-and-excitation networks, a channel attention mechanism, and concatenated attention neural network, a spatial attention mechanism, the model can more adeptly focus on the critical regions and finer details of brain tumors. Residual network-50 combined attention module and feature pyramid network as the backbone network to effectively capture multi-scale characteristics of brain tumors. At the same time, the region proposal network and region of interest align technology were used to ensure that the segmentation area matched the actual tumor morphology. The originality of the research lies in the deep residual network that combines attention mechanism with feature pyramid network to replace the backbone based on mask region convolutional neural network, achieving an improvement in the efficiency of brain tumor feature extraction. After a series of experiments, the precision of the model is 90.72%, which is 0.76% higher than that of the original model. Recall was 91.68%, an increase of 0.95%; Mean Intersection over Union was 94.56%, an increase of 1.39%. This method achieves precise segmentation of brain tumor magnetic resonance imaging, and doctors can easily and accurately locate the tumor area through the segmentation results, thereby quickly measuring the diameter, area, and other information of the tumor, providing doctors with more comprehensive diagnostic information.

SFA-Net: Semantic Feature Adjustment Network for Remote Sensing Image Segmentation

Advances in deep learning and computer vision techniques have made impacts in the field of remote sensing, enabling efficient data analysis for applications such as land cover classification and change detection. Convolutional neural networks (CNNs) and transformer architectures have been utilized in visual perception algorithms due to their effectiveness in analyzing local features and global context. In this paper, we propose a hybrid transformer architecture that consists of a CNN-based encoder and transformer-based decoder. We propose a feature adjustment module that refines the multiscale feature maps extracted from an EfficientNet backbone network. The adjusted feature maps are integrated into the transformer-based decoder to perform the semantic segmentation of the remote sensing images. This paper refers to the proposed encoder–decoder architecture as a semantic feature adjustment network (SFA-Net). To demonstrate the effectiveness of the SFA-Net, experiments were thoroughly conducted with four public benchmark datasets, including the UAVid, ISPRS Potsdam, ISPRS Vaihingen, and LoveDA datasets. The proposed model achieved state-of-the-art accuracy on the UAVid, ISPRS Vaihingen, and LoveDA datasets for the segmentation of the remote sensing images. On the ISPRS Potsdam dataset, our method achieved comparable accuracy to the latest model while reducing the number of trainable parameters from 113.8 M to 10.7 M.

V-YOLO: A Lightweight and Efficient Detection Model for Guava in Complex Orchard Environments

The global agriculture industry is encountering challenges due to labor shortages and the demand for increased efficiency. Currently, fruit yield estimation in guava orchards primarily depends on manual counting. Machine vision is an essential technology for enabling automatic yield estimation in guava production. To address the detection of guava in complex natural environments, this paper proposes an improved lightweight and efficient detection model, V-YOLO (VanillaNet-YOLO). By utilizing the more lightweight and efficient VanillaNet as the backbone network and modifying the head part of the model, we enhance detection accuracy, reduce the number of model parameters, and improve detection speed. Experimental results demonstrate that V-YOLO and YOLOv10n achieve the same mean average precision (mAP) of 95.0%, but V-YOLO uses only 43.2% of the parameters required by YOLOv10n, performs calculations at 41.4% of the computational cost, and exhibits a detection speed that is 2.67 times that of YOLOv10n. These findings indicate that V-YOLO can be employed for rapid detection and counting of guava, providing an effective method for visually estimating fruit yield in guava orchards.

An Empirical Investigation into an Economic Analysis of State Road Transport Undertakings in India

State Road Transport Undertakings have been the backbone of the transportation network in India since their inception. These corporations were set up to provide a sustainable and cheap mode of transport for citizens of India, along with providing connectivity to rural parts of the nation. Over the years, the financial and physical performance of most State Road Transport Undertakings has dropped significantly. The social goal, political influence and poor management have been some of the major factors contributing to the losses made by most of the State Road Transport Undertakings. We have conducted this study by considering eight State Road Transport Undertakings, which are loss and profit-making, to understand the factors responsible for improving the physical factors of these State Road Transport Undertakings. To undertake this objective, we used correlation analysis and feature importance analysis using random forest regressor. We have proposed some policy-based measures for each State Road Transport Undertaking analysed based on the results and analysis. JEL Codes: C12, I38, L32, L91, N35, O53, R41, R42

Spatial awareness enhancement based single-stage anchor-free 3D object detection for autonomous driving

The real-time and accurate detection of three-dimensional (3D) objects based on LiDAR is a focal problem in the field of autonomous driving environment perception. Compared to two-stage and anchor-based 3D object detection methods that suffer from inference latency challenges, single-stage anchor-free 3D object detection approaches are more suitable for deployment in autonomous driving vehicles with the strict real-time requirement. However, they face the issue of insufficient spatial awareness, which can result in detection errors such as false positives and false negatives, thereby increasing the potential risks of autonomous driving. In response to this, we focus on enhancing the spatial awareness of CenterPoint, a widely used single-stage anchor-free 3D object detector in the industry. Considering the limited allocation of computational resources and the performance bottleneck caused by pillar encoder, we propose an efficient SSDCM backbone to strengthen feature representation and extraction. Furthermore, a simple BGC neck is devised to weight and exchange contextual information in order to deeply fuse multi-scale features. Combining improved backbone and neck networks, we construct a single-stage anchor-free 3D object detection model with spatial awareness enhancement, named CenterPoint-Spatial Awareness Enhancement (CenterPoint-SAE). We evaluate CenterPoint-SAE on two large-scale and challenging autonomous driving datasets, nuScenes and Waymo. It achieves 53.3% mAP and 62.5% NDS on nuScenes detection benchmark, and runs inference at a speed of 11.1 FPS. Compared to the baseline, the upgraded networks deliver a performance improvement of 1.6% mAP and 1.2% NDS at minor cost. Notably, on Waymo dataset, our method achieves competitive detection performance compared to two-stage and point-based methods.

Insulator Defect Detection Based on YOLOv5s-KE

To tackle the issue of low detection accuracy in insulator images caused by intricate backgrounds and small defect sizes, as well as the requirement for real-time detection on embedded and mobile devices, this research introduces the YOLOv5s-KE model. Integrating multiple strategies, YOLOv5s-KE aims to boost detection accuracy significantly. Initially, an enhanced anchor generation method utilizing the K-means++ algorithm is proposed to generate more appropriate anchor boxes for insulator defects. Moreover, an attention mechanism is integrated into both the backbone and neck networks to enhance the model’s capacity to focus on defect features and resist interference. To improve the detection of small defects, the EIoU loss function is implemented in place of the original CIoU loss function. In order to meet the real-time detection needs on embedded and mobile devices, the model is further refined through the integration of Ghost convolution for lightweight feature extraction and a linear transformation to reduce the computational burden of standard convolution. A channel pruning strategy is deployed to optimize the sparsely trained network, diminishing redundancy, and improving model generalization. Additionally, the CARAFE operator replaces the original upsampling operator to minimize model parameters and elevate detection speed. Experimental outcomes demonstrate that YOLOv5s-KE achieves a detection accuracy of 92.3% on the Chinese transmission line insulator dataset, marking a 5.2% enhancement over the original YOLOv5s. The streamlined version of YOLOv5s-KE achieves a detection speed of 94.3 frames per second, indicating an improvement of 30.1 frames per second compared to the original model. Model parameters are condensed to 9.6 M, resulting in a detection accuracy of 91.1%. This study underscores the precision and efficiency of the proposed approach, suggesting that the advanced strategies explored introduce novel possibilities for insulator defect detection.

A Deformable and Multi-Scale Network with Self-Attentive Feature Fusion for SAR Ship Classification

The identification of ships in Synthetic Aperture Radar (SAR) imagery is critical for effective maritime surveillance. The advent of deep learning has significantly improved the accuracy of SAR ship classification and recognition. However, distinguishing features between different ship categories in SAR images remains a challenge, particularly as the number of categories increases. The key to achieving high recognition accuracy lies in effectively extracting and utilizing discriminative features. To address this, we propose DCN-MSFF-TR, a novel recognition model inspired by the Transformer encoder–decoder architecture. Our approach integrates a deformable convolutional module (DCN) within the backbone network to enhance feature extraction. Additionally, we introduce multi-scale self-attention processing from the Transformer into the feature hierarchy and fuse these representations at appropriate levels using a feature pyramid strategy. This enables each layer to leverage both its own information and synthesized features from other layers, enhancing feature representation. Extensive evaluations on the OpenSARShip-3-Complex and OpenSARShip-6-Complex datasets demonstrate the effectiveness of our method. DCN-MSFF-TR achieves average recognition accuracies of 78.1% and 66.7% on the three-class and six-class datasets, respectively, outperforming existing recognition models and showcasing its superior capability in accurately identifying ship categories in SAR images.

LSR-Det: A Lightweight Detector for Ship Detection in SAR Images Based on Oriented Bounding Box

Convolutional neural networks (CNNs) have significantly advanced in recent years in detecting arbitrary-oriented ships in synthetic aperture radar (SAR) images. However, challenges remain with multi-scale target detection and deployment on satellite-based platforms due to the extensive model parameters and high computational complexity. To address these issues, we propose a lightweight method for arbitrary-oriented ship detection in SAR images, named LSR-Det. Specifically, we introduce a lightweight backbone network based on contour guidance, which reduces the number of parameters while maintaining excellent feature extraction capability. Additionally, a lightweight adaptive feature pyramid network is designed to enhance the fusion capability of the ship features across different layers with a low computational cost by incorporating adaptive ship feature fusion modules between the feature layers. To efficiently utilize the fused features, a lightweight rotating detection head is designed, incorporating the idea of sharing the convolutional parameters, thereby improving the network’s ability to detect multi-scale ship targets. The experiments conducted on the SAR ship detection dataset (SSDD) and the rotating ship detection dataset (RSDD-SAR) demonstrate that LSR-Det achieves an average precision (AP50) of 98.5% and 97.2% with 3.21 G floating point operations (FLOPs) and 0.98 M parameters, respectively, outperforming the current popular SAR arbitrary-direction ship target detection methods.

Tiny-Object Detection Based on Optimized YOLO-CSQ for Accurate Drone Detection in Wildfire Scenarios

Wildfires, which are distinguished by their destructive nature and challenging suppression, present a significant threat to ecological environments and socioeconomic systems. In order to address this issue, the development of efficient and accurate fire detection technologies for early warning and timely response is essential. This paper addresses the complexity of forest and mountain fire detection by proposing YOLO-CSQ, a drone-based fire detection method built upon an improved YOLOv8 algorithm. Firstly, we introduce the CBAM attention mechanism, which enhances the model’s multi-scale fire feature extraction capabilities by adaptively adjusting weights in both the channel and spatial dimensions of feature maps, thereby improving detection accuracy. Secondly, we propose an improved ShuffleNetV2 backbone network structure, which significantly reduces the model’s parameter count and computational complexity while maintaining feature extraction capabilities. This results in a more lightweight and efficient model. Thirdly, to address the challenges of varying fire scales and numerous weak emission targets in mountain fires, we propose a Quadrupled-ASFF detection head for weighted feature fusion. This enhances the model’s robustness in detecting targets of different scales. Finally, we introduce the WIoU loss function to replace the traditional CIoU object detection loss function, thereby enhancing the model’s localization accuracy. The experimental results demonstrate that the improved model achieves an mAP@50 of 96.87%, which is superior to the original YOLOV8, YOLOV9, and YOLOV10 by 10.9, 11.66, and 13.33 percentage points, respectively. Moreover, it exhibits significant advantages over other classic algorithms in key evaluation metrics such as precision, recall, and F1 score. These findings validate the effectiveness of the improved model in mountain fire detection scenarios, offering a novel solution for early warning and intelligent monitoring of mountain wildfires.

Style Optimization Networks for real-time semantic segmentation of rainy and foggy weather

Semantic segmentation is an essential task in the field of computer vision. Existing semantic segmentation models can achieve good results under good weather and lighting conditions. However, when the external environment changes, the effectiveness of these models are seriously affected. Therefore, we focus on the task of semantic segmentation in rainy and foggy weather. Fog is a common phenomenon in rainy weather conditions and has a negative impact on image visibility. Besides, to make the algorithm satisfy the application requirements of mobile devices, the computational cost and the real-time requirement of the model have become one of the major points of our research. In this paper, we propose a novel Style Optimization Network (SONet) architecture, containing a Style Optimization Module (SOM) that can dynamically learn style information, and a Key information Extraction Module (KEM) that extracts important spatial and contextual information. This can improve the learning ability and robustness of the model for rainy and foggy conditions. Meanwhile, we achieve real-time performance by using lightweight modules and a backbone network with low computational complexity. To validate the effectiveness of our SONet, we synthesized CityScapes dataset for rainy and foggy weather and evaluated the accuracy and complexity of our model. Our model achieves a segmentation accuracy of 75.29% MIoU and 83.62% MPA on a NVIDIA TITAN Xp GPU. Several comparative experiments have shown that our SONet can achieve good performance in semantic segmentation tasks under rainy and foggy weather, and due to the lightweight design of the model we have a good advantage in both accuracy and model complexity.