MFANet: Multi-scale feature fusion network with attention mechanism

Wang, Gaihua; Xin, Gan; Cao, Qingcheng; Zhai, Qianyu

doi:10.1007/s00371-022-02503-4

Cited by 13 publications

(5 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multiscale feature fusion networks [15] play a key role in target detection, helping models capture target details and context information at different scales to improve model detection performance. In summary, the role of deep learning-based feature fusion networks in target detection is to enhance the model's understanding of different scales, types of information, and contexts, thereby improving the accuracy and robustness of target detection.…”

Section: 2mentioning

confidence: 99%

FEM-Net: Industrial Surface Defect Detection Network Based on Feature Enhancement and Efficient Multiscale Feature Fusion

duan,

Wang,

Zhao

2024

Preprint

View full text Add to dashboard Cite

Industrial defect detection methods based on deep learning can reduce the cost of traditional manual quality checks and improve the accuracy and efficiency of testing. However, in industrial inspection scenarios, problems such as complex industrial image texture backgrounds, low contrast, and significant changes in defect size challenge existing industrial inspection networks. To solve the above problems, an industrial surface defect detection network (FEM-Net) based on feature enhancement and efficient multiscale feature fusion is proposed. Firstly, a new lightweight feature extraction network is designed in this paper. (LFNNet) to reduce the number of parameters and calculation costs of the model, while introducing a cascading channel space attention module (CSM) to pay more attention to the location of defective features of industrial products. Secondly, to improve the texture background complex small target detection and enhance the characterization of the target feature, a multi-level feature enhancement module (MFEM) is proposed to fully capture the target object’s context information and global relationship. Finally, an efficient multiscale feature fusion network (EMF-Net) is designed to effectively blend the characteristics of each layer extracted from the backbone network. Use the Adaptive Weighted Feature Fusion Module (AFM) to integrate features at multiple adjacent levels before entering the detector to mitigate fusion differences between multiscale features. The experimental results show that the FEM-Net network obtained 94.4% and 98.8% mAP @.5 on the NEU steel and PCB datasets, respectively, compared with the mainstream target detector.

show abstract

Section: 2mentioning

confidence: 99%

FEM-Net: Industrial Surface Defect Detection Network Based on Feature Enhancement and Efficient Multiscale Feature Fusion

duan,

Wang,

Zhao

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…However, deep learning algorithms require a sufficient number of samples. Using augmented image methods excessively for quantity expansion can result in the repetition of image features, which leads to overfitting models [38].…”

Section: Dataset Constructionmentioning

confidence: 99%

“…Unlike conventional target recognition, wildlife recognition in complex forest environments is a challenge. This is due to several factors, such as dense tree growth, unpredictable weather conditions, moving shadows, and distractions like rain and fog [36]. Additionally, the natural camouflage of wild animals further complicates their identification in these environments (Figure 1) [37].…”

Section: Introductionmentioning

confidence: 99%

Wildlife Real-Time Detection in Complex Forest Scenes Based on YOLOv5s Deep Learning Network

Ma,

Dong,

Xia

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

With the progressively deteriorating global ecological environment and the gradual escalation of human activities, the survival of wildlife has been severely impacted. Hence, a rapid, precise, and reliable method for detecting wildlife holds immense significance in safeguarding their existence and monitoring their status. However, due to the rare and concealed nature of wildlife activities, the existing wildlife detection methods face limitations in efficiently extracting features during real-time monitoring in complex forest environments. These models exhibit drawbacks such as slow speed and low accuracy. Therefore, we propose a novel real-time monitoring model called WL-YOLO, which is designed for lightweight wildlife detection in complex forest environments. This model is built upon the deep learning model YOLOv5s. In WL-YOLO, we introduce a novel and lightweight feature extraction module. This module is comprised of a deeply separable convolutional neural network integrated with compression and excitation modules in the backbone network. This design is aimed at reducing the number of model parameters and computational requirements, while simultaneously enhancing the feature representation of the network. Additionally, we introduced a CBAM attention mechanism to enhance the extraction of local key features, resulting in improved performance of WL-YOLO in the natural environment where wildlife has high concealment and complexity. This model achieved a mean accuracy (mAP) value of 97.25%, an F1-score value of 95.65%, and an accuracy value of 95.14%. These results demonstrated that this model outperforms the current mainstream deep learning models. Additionally, compared to the YOLOv5m base model, WL-YOLO reduces the number of parameters by 44.73% and shortens the detection time by 58%. This study offers technical support for detecting and protecting wildlife in intricate environments by introducing a highly efficient and advanced wildlife detection model.

show abstract

“…In 2D image tasks, Refs. 31 to 34 used a pyramid structure to fuse multiscale feature information with good results. In 3D point clouds, Liang et al 35 .…”

Section: Related Workmentioning

confidence: 99%

Adaptive offset self-attention network for 3D point cloud

Wang

et al. 2023

J. Electron. Imag.

View full text Add to dashboard Cite

.The semantic segmentation of point clouds has achieved significant progress in recent years. However, most current methods rely on complex stacking of modules to learn local features, resulting in a complicated network structure. We propose an adaptive offset self-attention network (AOSANet), a simple and efficient point cloud learning framework for classification and segmentation tasks. AOSANet has permutation invariance when processing three-dimensional point cloud data and can dynamically learn local features based on the input features. To better fuse the local features, we introduce the maximum feature extraction module, which performs maximum feature weighting prior to local feature fusion. This allows the network to concentrate on the most important features and improve overall robustness. Through extensive experiments on publicly available datasets, the proposed method achieves a state-of-the-art performance on point cloud classification and segmentation tasks.

show abstract

MFANet: Multi-scale feature fusion network with attention mechanism

Cited by 13 publications

References 33 publications

FEM-Net: Industrial Surface Defect Detection Network Based on Feature Enhancement and Efficient Multiscale Feature Fusion

FEM-Net: Industrial Surface Defect Detection Network Based on Feature Enhancement and Efficient Multiscale Feature Fusion

Wildlife Real-Time Detection in Complex Forest Scenes Based on YOLOv5s Deep Learning Network

Adaptive offset self-attention network for 3D point cloud

Contact Info

Product

Resources

About