High-level Semantic Information Research Articles

Building footprint extraction from aerial images using an edge-aware YOLO-v8 network

Abstract. Building footprint extraction is a critical indicator for assessing urban infrastructure, and extracting building footprints from remote sensing imagery can have significant practical applications. However, achieving rapid and accurate extraction of building footprints remains highly challenging, especially in scenarios with complex scenes, dense building distributions, and small targets. The instance segmentation models of the YOLO series offer strong real-time performance, reducing considerable time and effort in practical applications. Therefore, we propose building footprint extraction based on an enhance YOLO-v8 network. This study focuses on three enhancements to the YOLO-v8 network to improve extraction accuracy. Building upon the YOLO-v8 framework, we have incorporated the Feature Pyramid Network (FPN) module into feature maps at all scales to efficiently propagate high-level semantic information. Additionally, we introduce the Triple Feature Encoder (TFE) module, which integrates spatial detail information from feature maps at three different scales to enhance the network's ability to extract multi-scale information. Finally, we explore the integration of the Prewitt model, a conventional edge detection operator, to assist in extracting edge features in target regions of feature maps. This integration aims to reduce the jagged edges frequently seen in the outcomes of the original YOLO-v8. Furthermore, the Prewitt operator's noise suppression capability helps mitigate the influence of non-target areas in the feature maps. The proposed framework achieves an instance segmentation accuracy of mAP50 is 84.6% and mAP50:95 is 51.4% on public datasets, outperforming the original YOLO-v8 network.

HST-MRF: Heterogeneous Swin Transformer With Multi-Receptive Field for Medical Image Segmentation.

The Transformer has been successfully used in medical image segmentation due to its excellent long-range modeling capabilities. However, patch segmentation is necessary when building a Transformer class model. This process ignores the tissue structure features within patch, resulting in the loss of shallow representation information. In this study, we propose a Heterogeneous Swin Transformer with Multi-Receptive Field (HST-MRF) model that fuses patch information from different receptive fields to solve the problem of loss of feature information caused by patch segmentation. The heterogeneous Swin Transformer (HST) is the core module, which achieves the interaction of multi-receptive field patch information through heterogeneous attention and passes it to the next stage for progressive learning, thus complementing the patch structure information. We also designed a two-stage fusion module, multimodal bilinear pooling (MBP), to assist HST in further fusing multi-receptive field information and combining low-level and high-level semantic information for accurate localization of lesion regions. In addition, we developed adaptive patch embedding (APE) and soft channel attention (SCA) modules to retain more valuable information when acquiring patch embedding and filtering channel features, respectively, thereby improving model segmentation quality. We evaluated HST-MRF on multiple datasets for polyp, skin lesion and breast ultrasound segmentation tasks. Experimental results show that our proposed method outperforms state-of-the-art models and can achieve superior performance. Furthermore, we verified the effectiveness of each module and the benefits of multi-receptive field segmentation in reducing the loss of structural information through ablation experiments and qualitative analysis.

Robust coverless image steganography based on human pose estimation

Coverless image steganography has garnered significant attention in the field of information hiding due to its robust anti-stego analysis security, as it eliminates the need for modifying the carrier image. With the proliferation of social platforms, a plethora of images containing abundant human information have emerged on the Internet, rendering them an ideal natural image carrier database. In this paper, we propose a pose estimation network model and devise a coverless image steganography algorithm based on pose estimation to conceal confidential information by mapping it onto human pose data. Since all selected carrier images are human-centered and statistically insensitive, and the high-level semantic information shows excellent stability against attacks, the concealment and robustness are enhanced. Experimental results on three public datasets demonstrate that the algorithm achieves an average robustness of up to 92% against multiple attacks.

Bi-directional information guidance network for UAV vehicle detection

AbstractUAV vehicle detection based on convolutional neural network exits a key problem of information imbalance of different feature layers. Shallow features have spatial information that is beneficial to localization, but lack semantic information. On the contrary, deep features have semantic information that is beneficial to classification, but lack spatial information. However, accurate classification and localization for UAV vehicle detection require both shallow spatial information and high semantic information. In our work, a bi-directional information guidance network (BDIG-Net) for UAV vehicle detection is proposed, which can ensure that each feature prediction layer has abundant mid-/low-level spatial information and high-level semantic information. There are two main parts in the BDIG-Net: shallow-level spatial information guidance part and deep-level semantic information guidance part. In the shallow-level guidance part, we design a feature transform module (FTM) to supply the mid-/low-level feature information, which can guide the BDIG-Net to enhance detailed and spatial features for deep features. Furthermore, we adopt a light-weight attention module (LAM) to reduce unnecessary shallow background information, making the network more focused on small-sized vehicles. In the deep-level guidance part, we use classical feature pyramid network to supply high-level semantic information, which can guide the BDIG-Net to enhance contextual information for shallow features. Meanwhile, we design a feature enhancement module (FEM) to suppress redundant features and improve the discriminability of vehicles. The proposed BDIG-Net can reduce the information imbalance. The experimental results show that the BDIG-Net can achieve accurate classification and localization for UAV vehicles and realize the real-time application requirements.

Towards interpretability lightweight semantic segmentation model for waterbody extraction in large-scale high resolution remote sensing images

ABSTRACT In recent years, high resolution (HR) remote sensing images have brought significant changes to land cover monitoring. Monitoring water resources, a vital and scarce commodity for human survival, is a crucial aspect of land cover monitoring and forms the foundation for water resource allocation in regions facing shortages. Hence, a universal method is necessary to address this problem and handle large-scale water resources monitoring. We explore the network structure via improving the interpretability to reduce the network layers and propose a lightweight pixel-wise semantic segmentation model, which achieves waterbody extraction from GF-1 remote sensing images with high accuracy and high speed. The proposed model structure addresses both semantic segmentation for large and small waterbodies, extracting fine-grained edge and contour information from high-resolution feature maps. Additionally, it efficiently extracts multi-scale high-level semantic information using residual convolutional blocks and dilated convolutional blocks. The multi-scale feature maps are fused, and binary classification is predicted through Support Vector Machines (SVM). Furthermore, the paper introduces a model training method with an adaptive learning rate, reducing the overall training time. To validate the model’s performance, remote sensing images from GF-1 are utilized to construct a dataset. Experimental results, compared with five models, demonstrate that the proposed method achieves the highest accuracy and least training time.

Research on Computer Classification Algorithm of Concrete Crack Based on Deep Learning

In combination with transfer learning under the framework of Unet semantic partitioning, VGG16 pre-trained neural network enhanced encoder is used to extract multi-level high-level semantic information. The cross-entropy loss function is used to eliminate the imbalance between samples, and finally the crack shape is accurately semitone. Combining with the theory of computer vision, a quantitative calculation method of the crack region, length, width and other geometric characteristic parameters based on the binary segmentation template is proposed. Finally, the self-developed dam concrete crack image is taken as an example to carry out simulation and comparison test to check the correctness and superiority of the research results of this project. The research results will reveal that the crack recognition based on deep neural network can achieve a high recognition rate. The calculation results of fracture characteristic parameters meet the requirement of detection accuracy. The research results of this paper are expected to provide a new technical means for the quality control of dam concrete structure.

Real-time Attentive Dilated U-Net for Extremely Dark Image Enhancement

Images taken under low-light conditions suffer from poor visibility, color distortion and graininess, all of which degrade the image quality and hamper the performance of downstream vision tasks, such as object detection and instance segmentation in the field of autonomous driving, making low-light enhancement an indispensable basic component of high-level visual tasks. Low-light enhancement aims to mitigate these issues, and has garnered extensive attention and research over several decades. The primary challenge in low-light image enhancement arises from the low signal-to-noise ratio (SNR) caused by insufficient lighting. This challenge becomes even more pronounced in near-zero lux conditions, where noise overwhelms the available image information. Both traditional image signal processing (ISP) pipeline and conventional low-light image enhancement methods struggle in such scenarios. Recently, deep neural networks have been used to address this challenge. These networks take unmodified RAW images as input and produce the enhanced sRGB images, forming a deep learning-based ISP pipeline. However, most of these networks are computationally expensive and thus far from practical use. In this paper, we propose a lightweight model called attentive dilated U-Net (ADU-Net) to tackle this issue. Our model incorporates several innovative designs, including an asymmetric U-shape architecture, dilated residual modules (DRMs) for feature extraction, and attentive fusion modules (AFMs) for feature fusion. The DRMs provide strong representative capability while the AFMs effectively leverage low-level texture information and high-level semantic information within the network. Both modules employ a lightweight design but offer significant performance gains. Extensive experiments demonstrate our method is highly-effective, achieving an excellent balance between image quality and computational complexity, i . e ., taking less than 4ms for a high-definition 4K image on a single GTX 1080Ti GPU and yet maintaining competitive visual quality. Furthermore, our method exhibits pleasing scalability and generalizability, highlighting its potential for widespread applicability.

HRD-Net: High resolution segmentation network with adaptive learning ability of retinal vessel features

Retinal segmentation is a crucial step in the early warning of human health conditions. However, retinal blood vessels possess complex curvature, irregular distribution, and contain multi-scale fine structures, which make the limited receptive field of regular convolution challenging to process their vascular details efficiently. Additionally, the encoder-decoder based network leads to irreversible spatial information loss because of multiple downsampling, resulting in over-segmentation and missed segmentation of the vessels. For this reason, we develop a high-resolution network based on Deformable Convolution v3, called HRD-Net. By constructing a high-resolution representation, the network allows special attention to be paid to the details of tiny blood vessels. The proposed feature enhancement cascade module based on Deformable Convolution v3 can flexibly adapt and capture the ever-changing morphology and intricate connections of retinal blood vessels, ensuring the continuity of vessel segmentation. In the output phase of the network, the proposed global aggregation module integrates full-resolution feature maps while suppressing redundant features, achieving an effective fusion of high-level semantic information and spatial detail information. In addition, we have re-examined the selection criteria for activation and normalization methods, and also refine the network architectures from a spatial domain perspective to release redundant computational loads. Testing on the DRIVE, STARE, and CHASE_DB1 datasets indicates that HRD-Net, with fewer parameters, outperforms existing segmentation methods on several evaluation metrics such as F1, ACC, SE, SP, AUC, and IOU.

CHL-DTI: A Novel High-Low Order Information Convergence Framework for Effective Drug-Target Interaction Prediction.

Recognizing drug-target interactions (DTI) stands as a pivotal element in the expansive field of drug discovery. Traditional biological wet experiments, although valuable, are time-consuming and costly as methods. Recently, computational methods grounded in network learning have demonstrated great advantages by effective topological feature extraction and attracted extensive research attention. However, most existing network-based learning methods only consider the low-order binary correlation between individual drug and target, neglecting the potential higher-order correlation information derived from multiple drugs and targets. High-order information, as an essential component, exhibits complementarity with low-order information. Hence, the incorporation of higher-order associations between drugs and targets, while adequately integrating them with the existing lower-order information, could potentially yield substantial breakthroughs in predicting drug-target interactions. We propose a novel dual channels network-based learning model CHL-DTI that converges high-order information from hypergraphs and low-order information from ordinary graph for drug-target interaction prediction. The convergence of high-loworder information in CHL-DTI is manifested in two key aspects. First, during the feature extraction stage, the model integrates both high-level semantic information and low-level topological information by combining hypergraphs and ordinary graph. Second, CHL-DTI fully fuse the innovative introduced drug-protein pairs (DPP) hypergraph network structure with ordinary topological network structure information. Extensive experimentation conducted on three public datasets showcases the superior performance of CHL-DTI in DTI prediction tasks when compared to SOTA methods. The source code of CHL-DTI is available at https://github.com/UPCLyy/CHL-DTI .

SC-Net: Symmetrical conical network for colorectal pathology image segmentation

Background and ObjectiveImage segmentation of histopathology of colorectal cancer is a core task of computer aided medical image diagnosis system. Existing convolutional neural networks generally extract multi-scale information in linear flow structures by inserting multi-branch modules, which is difficult to extract heterogeneous semantic information under multi-level and different receptive field and tough to establish context dependency among different receptive field features. MethodsTo address these issues, we propose a symmetric spiral progressive feature fusion encoder-decoder network called the Symmetric Conical Network (SC-Net). First, we design a Multi-scale Feature Extraction Block (MFEB) matching with the Symmetric Conical Network to obtain multi-branch heterogeneous semantic information under different receptive fields, so as to enrich the diversity of extracted feature information. The encoder is composed of MFEB through spiral and multi-branch arrangement to enhance context dependence between different information flow. Secondly, the information loss of contour, color and others in high-level semantic information through causally stacking MFEB, the Feature Mapping Layer (FML) is designed to map low-level features to high-level semantic features along the down-sampling branch and solve the problem of insufficient global feature extraction in deep levels. ResultsThe SC-Net was evaluated on our self-constructed colorectal cancer dataset, a publicly available breast cancer dataset and a polyp dataset. The results revealed that the mDice of segmentation reached 0.8611, 0.7259 and 0.7144. We compare our model with the state-of-art semantic segmentation UNet++, PSPNet, Attention U-Net, R2U-Net and other advanced segmentation networks. The experimental results demonstrate that we achieve the most advanced performance. ConclusionsThe results indicate that the proposed SC-Net excels in segmenting H&E stained pathology images, effectively preserving morphological features and spatial information even in scenarios with weak texture, poor contrast, and variations in appearance.

Entropy Kernel Graph Cut Feature Space Enhancement with SqueezeNet Deep Neural Network for Textural Image Segmentation

Recently, image segmentation based on graph cut methods has shown remarkable performance on a set of image data. Although the kernel graph cut method provides good performance, its performance is highly dependent on the data mapping to the transformation space and image features. The entropy-based kernel graph cut method is suitable for segmentation of textured images. Nonetheless, its segmentation quality remains significantly contingent on the accuracy and richness of feature space representation and kernel centers. This paper introduces an entropy-based kernel graph cut method, which leverages the discriminative feature space extracted from SqueezeNet, a deep neural network. The fusion of SqueezeNet’s features enriches the segmentation process by capturing high-level semantic information. Moreover, the extraction of kernel centers is refined through a weighted k-means approach, contributing further to the segmentation’s precision and effectiveness. The proposed method, while exploiting the benefits of suitable computational load of graph cut methods, will be a suitable alternative for segmenting textured images. Laboratory results have been taken on a set of well-known datasets that include textured shapes in order to evaluate the efficiency of the algorithm compared to other well-known methods in the field of kernel graph cut.

Multi-label text classification of cardiovascular drug attributes based on BERT and BiGRU

Extracting and digitizing drug attributes from medical literature is the first step to build a knowledge computing system for precision disease treatment. In order to build a cardiovascular drug knowledge base, this paper proposes a multi-label text classification method for cardiovascular drug attributes from the Chinese drug guideline. The drug attributes are characterized by a BERT pre-trained model, and a dual-feature extraction structure is proposed based on the BiGRU neural network to capture high-level semantic information. Label categorization of cardiovascular drug attributes, such as indications and mode of administration, is accomplished. The F1 score of 0.8431 was obtained using 5-fold cross-validation. Comparing KNN and Naïve bayes, and conducting CNN and BiGRU control experiments on the basis of Word2Vec characterization of medication guidelines, the proposed multi-label text classification method is effective and the F1 value is significantly improved. Proved by analysis of ablation and crossover experiments, the proposed method can achieve a high accuracy rate averaged at 0.8339.

Bidirectional Feature Aggregation Network for Stereo Image Quality Assessment Considering Parallax Attention-Based Binocular Fusion

Inspired by the two-path visual information processing mechanism (i.e., a bottom-up path and a top-down path), we propose a bidirectional binocular feature aggregation based stereo image quality assessment (SIQA) network, which considers a two-path visual mechanism and realizes the binocular fusion based on parallax information. To better aggregate binocular features from different levels, a two-path feature aggregation structure, which simulates the bottom-up and top-down mechanism in human visual system (HVS), is proposed. It not only realizes the supplement of low-level detail information to high-level semantic in the bottom-up path, but also realizes the supplement of high-level semantic information to low-level detail in the top-down path. Simultaneously, because feature misalignment exists in binocular features of adjacent levels, a feature alignment module (FAM) based on deformable convolution is designed to integrate the binocular fusion features of adjacent levels. In addition, considering the importance role of parallax in guiding binocular fusion, a binocular fusion module (BFM) based on parallax attention mechanism, which is different with existing binocular fusion methods, is explicitly proposed to achieve the binocular fusion between the left and right view features. Extensive experiments are conducted on LIVE I, LIVE II, WIVC I and WIVC II databases to demonstrate the effectiveness of the proposed method.

Multi-angle perception and convolutional neural network for service quality evaluation of cross-border e-commerce logistics enterprise.

The development of cross-border e-commerce logistics services has injected new vitality into the development of international trade, and therefore has become a new hot spot in theoretical research. In order to ensure the healthy development of cross-border e-commerce, it is urgent to build a set of scientific and effective evaluation mechanisms to scientifically evaluate the logistics service quality of cross-border e-commerce. Multi-angle perceptual convolutional neural network is a framework for service scene identification of cross-border e-commerce logistics enterprises based on deep convolutional neural network and multi-angle perceptual width learning. In this article, both shallow features and deep features were input into the deep perception model (DPM) to obtain a set of distinguishable features with causal structure, which was used to completely describe the high-level semantic information of cross-border e-commerce logistics enterprise services. Among them, DPM mainly adopts the fusion strategy of shallow feature and deep feature. Meanwhile, the feature representation is input into the width learning pattern recognition system for training and classification, so as to evaluate the service quality of cross-border e-commerce logistics enterprises. The multi-angle perceptual convolutional neural network can effectively solve the problems of high similarity between service classes of cross-border e-commerce logistics enterprises and large differences within the class, and achieve better generalization performance and algorithm complexity than support vector machine, random forest and convolutional neural network.

A novel deep-learning based weighted feature fusion architecture for precise classification of pressure injury.

Introduction: Precise classification has an important role in treatment of pressure injury (PI), while current machine-learning or deeplearning based methods of PI classification remain low accuracy. Methods: In this study, we developed a deeplearning based weighted feature fusion architecture for fine-grained classification, which combines a top-down and bottom-up pathway to fuse high-level semantic information and low-level detail representation. We validated it in our established database that consist of 1,519 images from multi-center clinical cohorts. ResNeXt was set as the backbone network. Results: We increased the accuracy of stage 3 PI from 60.3% to 76.2% by adding weighted feature pyramid network (wFPN). The accuracy for stage 1, 2, 4 PI were 0.870, 0.788, and 0.845 respectively. We found the overall accuracy, precision, recall, and F1-score of our network were 0.815, 0.808, 0.816, and 0.811 respectively. The area under the receiver operating characteristic curve was 0.940. Conclusions: Compared with current reported study, our network significantly increased the overall accuracy from 75% to 81.5% and showed great performance in predicting each stage. Upon further validation, our study will pave the path to the clinical application of our network in PI management.

ReFID: Reciprocal Frequency-aware Generalizable Person Re-identification via Decomposition and Filtering

Domain generalization (DG) of person re-identification (ReID) aims to conduct testing across domains that have not been previously encountered, without utilizing target domain data during the training stage. As the number of source domains increases, the relationships between training samples become more complex. This can lead to domain-invariant features that include certain instance-level spurious correlations, which can impact the model’s ability to generalize further. To overcome this limitation, the Reciprocal Frequency-aware Generalizable Person Re-identification (ReFID) method is proposed in this paper, which aims to utilize spectral feature correlation learning to transmit frequency component information and generate more discriminative hybrid features. A module called Bilateral Frequency Component-guided Attention (BFCA) is developed to help the network understand high-level semantic and texture information from various frequency features. Furthermore, to reduce the impact of noise from the frequency domain, this paper proposes an innovative module called Fourier Noise Masquerade Filtering (FNMF). This module enhances the portability of frequency domain components while simultaneously suppressing elements that do not contribute to generalization. Extensive experimental results on various datasets demonstrate that our method is effective and superior to the state-of-the-art methods.

Knowledge Distillation via Token-Level Relationship Graph Based on the Big Data Technologies

In the big data era, characterized by vast volumes of complex data, the efficiency of machine learning models is of utmost importance, particularly in the context of intelligent agriculture. Knowledge distillation (KD), a technique aimed at both model compression and performance enhancement, serves as a pivotal solution by distilling the knowledge from an elaborate model (teacher) to a lightweight, compact counterpart (student). However, the true potential of KD has not been fully explored. Existing approaches primarily focus on transferring instance-level information by big data technologies, overlooking the valuable information embedded in token-level relationships, which may be particularly affected by the long-tail effects. To address the above limitations, we propose a novel method called Knowledge Distillation with Token-level Relationship Graph (TRG) that leverages token-wise relationships to enhance the performance of knowledge distillation. By employing TRG, the student model can effectively emulate higher-level semantic information from the teacher model, resulting in improved performance and mobile-friendly efficiency. To further enhance the learning process, we introduce a dynamic temperature adjustment strategy, which encourages the student model to capture the topology structure of the teacher model more effectively. We conduct experiments to evaluate the effectiveness of the proposed method against several state-of-the-art approaches. Empirical results demonstrate the superiority of TRG across various visual tasks, including those involving imbalanced data. Our method consistently outperforms the existing baselines, establishing a new state-of-the-art performance in the field of KD based on big data technologies.

ETUNet:Exploring efficient transformer enhanced UNet for 3D brain tumor segmentation

Medical image segmentation is a crucial topic in medical image processing. Accurately segmenting brain tumor regions from multimodal MRI scans is essential for clinical diagnosis and survival prediction. However, similar intensity distributions, variable tumor shapes, and fuzzy boundaries pose severe challenges for brain tumor segmentation. Traditional segmentation networks based on UNet struggle to establish explicit long-range dependencies from the feature space due to the limitations of the CNN receptive field. This is particularly crucial for dense prediction tasks such as brain tumor segmentation. Recent works have incorporated the powerful global modeling capability of Transformer into UNet to achieve more precise segmentation results. Nevertheless, these methods encounter some issues: (1) the global information is often modeled by simply stacking Transformer layers for a specific module, resulting in high computational complexity and underutilization of the potential of the UNet architecture; (2) the rich boundary information of tumor subregions in multi-scale features is often overlooked. Motivated by these challenges, we propose an advanced fusion of Transformer with UNet by reexamining the core three parts (encoder, bottleneck, and skip connections). Firstly, we introduce a CNN-Transformer module in the encoder to replace the traditional CNN module, enabling the capture of deep spatial dependencies from input images. To address high-level semantic information, we incorporate a computationally efficient spatial-channel attention layer in the bottleneck for global interaction, highlighting important semantic features from the encoder path output. For irregular lesions, we fuse the multi-scale features from the encoder output and the decoder features in the skip connections by calculating cross-attention. This adaptive querying of valuable information from multi-scale features enhances the boundary localization ability of the decoder path and suppresses redundant features with low correlation. Compared to existing methods, our model further enhances the learning capacity of the overall UNet architecture while maintaining low computational complexity. Experimental results on the BraTS2018 and BraTS2020 datasets for brain tumor segmentation tasks demonstrate that our model achieves comparable or superior results compared to recent CNN or Transformer-based models. The average DSC and HD95 on the two datasets are 0.854, 6.688, and 0.862, 5.455 respectively. At the same time, our model achieves optimal segmentation of Enhancing tumors, showcasing the effectiveness of our method. Our code will be made publicly available at https://github.com/wzhangck/ETUnet.

ResU-Former: Advancing Remote Sensing Image Segmentation with Swin Residual Transformer for Precise Global–Local Feature Recognition and Visual–Semantic Space Learning

In the field of remote sensing image segmentation, achieving high accuracy and efficiency in diverse and complex environments remains a challenge. Additionally, there is a notable imbalance between the underlying features and the high-level semantic information embedded within remote sensing images, and both global and local recognition improvements are also limited by the multi-scale remote sensing scenery and imbalanced class distribution. These challenges are further compounded by inaccurate local localization segmentation and the oversight of small-scale features. To achieve balance between visual space and semantic space, to increase both global and local recognition accuracy, and to enhance the flexibility of input scale features while supplementing global contextual information, in this paper, we propose a U-shaped hierarchical structure called ResU-Former. The incorporation of the Swin Residual Transformer block allows for the efficient segmentation of objects of varying sizes against complex backgrounds, a common scenario in remote sensing datasets. With the specially designed Swin Residual Transformer block as its fundamental unit, ResU-Former accomplishes the full utilization and evolution of information, and the maximum optimization of semantic segmentation in complex remote sensing scenarios. The standard experimental results on benchmark datasets such as Vaihingen, Overall Accuracy of 81.5%, etc., show the ResU-Former’s potential to improve segmentation tasks across various remote sensing applications.

HRYNet: A Highly Robust YOLO Network for Complex Road Traffic Object Detection.

Object detection is a crucial component of the perception system in autonomous driving. However, the road scene presents a highly intricate environment where the visibility and characteristics of traffic targets are susceptible to attenuation and loss due to various complex road scenarios such as lighting conditions, weather conditions, time of day, background elements, and traffic density. Nevertheless, the current object detection network must exhibit more learning capabilities when detecting such targets. This also exacerbates the loss of features during the feature extraction and fusion process, significantly compromising the network's detection performance on traffic targets. This paper presents a novel methodology by which to overcome the concerns above, namely HRYNet. Firstly, a dual fusion gradual pyramid structure (DFGPN) is introduced, which employs a two-stage gradient fusion strategy to enhance the generation of more comprehensive multi-scale high-level semantic information, strengthen the interconnection between non-adjacent feature layers, and reduce the information gap that exists between them. HRYNet introduces an anti-interference feature extraction module, the residual multi-head self-attention mechanism (RMA). RMA enhances the target information by implementing a characteristic channel weighting policy, thereby reducing background interference and improving the attention capability of the network. Finally, the detection performance of HRYNet was evaluated by utilizing three datasets: the horizontally collected dataset BDD1000K, the UAV high-altitude dataset Visdrone, and a custom dataset. Experimental results demonstrate that HRYNet achieves a higher mAP_0.5 compared with YOLOv8s on the three datasets, with increases of 10.8%, 16.7%, and 5.5%, respectively. To optimize HRYNet for mobile devices, this study presents Lightweight HRYNet (LHRYNet), which effectively reduces the number of model parameters by 2 million. The results demonstrate that LHRYNet outperforms YOLOv8s in terms of mAP_0.5, with improvements of 6.7%, 10.9%, and 2.5% observed on the three datasets, respectively.