Search

The advancement of new energy vehicles has led to more demanding standards for detecting defects in cylindrical coated lithium batteries. The current research lacks robustness and has low performance. This paper seeks to provide real-time defect identification in cylindrical coated lithium batteries and improve the object detection method of the YOLOv5s model. This paper presents an MGSEC3 module with multi-scale feature extraction and integration of the SENet network in the YOLOv5 Backbone network. This module aims to reduce computational burden and enhance feature extraction efficiency as much as possible. At the same time, the CARAFE operator is utilized to enhance the up-sampling operator in order to reduce the loss of feature information. In addition, enhancements to the loss function enhance both the detection performance and convergence speed. The enhanced YOLOv5s model achieved an average detection rate of 82.4% on the custom cylindrical coated lithium battery dataset, 2.3% higher than the original YOLOv5s. This paper significantly enhances the precision and efficiency of flaw identification in cylindrical coated lithium batteries. [ABSTRACT FROM AUTHOR]

Scene text detection is an important research field in computer vision, playing a crucial role in various application scenarios. However, existing scene text detection methods often fail to achieve satisfactory results when faced with text instances of different sizes, shapes, and complex backgrounds. To address the challenge of detecting diverse texts in natural scenes, this paper proposes a multi-scale natural scene text detection method based on attention feature extraction and cascaded feature fusion. This method combines global and local attention through an improved attention feature fusion module (DSAF) to capture text features of different scales, enhancing the network's perception of text regions and improving its feature extraction capabilities. Simultaneously, an improved cascaded feature fusion module (PFFM) is used to fully integrate the extracted feature maps, expanding the receptive field of features and enriching the expressive ability of the feature maps. Finally, to address the cascaded feature maps, a lightweight subspace attention module (SAM) is introduced to partition the concatenated feature maps into several sub-space feature maps, facilitating spatial information interaction among features of different scales. In this paper, comparative experiments are conducted on the ICDAR2015, Total-Text, and MSRA-TD500 datasets, and comparisons are made with some existing scene text detection methods. The results show that the proposed method achieves good performance in terms of accuracy, recall, and F-score, thus verifying its effectiveness and practicality. [ABSTRACT FROM AUTHOR]

Target detection algorithms can greatly improve the efficiency of tomato leaf disease detection and play an important technical role in intelligent tomato cultivation. However, there are some challenges in the detection process, such as the diversity of complex backgrounds and the loss of leaf symmetry due to leaf shadowing, and existing disease detection methods have some disadvantages in terms of deteriorating generalization ability and insufficient accuracy. Aiming at the above issues, a target detection model for tomato leaf disease based on deep learning with a global attention mechanism, TDGA, is proposed in this paper. The main idea of TDGA includes three aspects. Firstly, TDGA adds a global attention mechanism (GAM) after up-sampling and down-sampling, as well as in the SPPF module, to improve the feature extraction ability of the target object, effectively reducing the interference of invalid targets. Secondly, TDGA uses a switchable atrous convolution (SAConv) in the C3 module to improve the model's ability to detect. Thirdly, TDGA adopts the efficient IoU loss (EIoU) instead of complete IoU loss (CIoU) to solve the ambiguous definition of aspect ratio and sample imbalance. In addition, the influences of different environmental factors such as single leaf, multiple leaves, and shadows on the performance of tomato disease detection are extensively experimented with and analyzed in this paper, which also verified the robustness of TDGA. The experimental results show that the average accuracy of TDGA reaches 91.40%, which is 2.93% higher than that of the original YOLOv5 network, which is higher than YOLOv5, YOLOv7, YOLOHC, YOLOv8, SSD, Faster R-CNN, RetinaNet and other target detection networks, so that TDGA can be utilized for the detection of tomato leaf disease more efficiently and accurately, even in complex environments. [ABSTRACT FROM AUTHOR]

Accurate and comprehensive detection of pavement cracks is important for maintaining road quality and ensuring traffic safety. However, the complexity of road surfaces and the diversity of cracks make it difficult for existing methods to accomplish this challenging task. This paper proposes a novel network named the global graph multiscale network (GGMNet) for automated pixel-level detection of pavement cracks. The GGMNet network has several innovations compared with the mainstream road crack detection network: (1) a global contextual Res-block (GC-Resblock) is proposed to guide the network to emphasize the identities of cracks while suppressing background noises; (2) a graph pyramid pooling module (GPPM) is designed to aggregate the multi-scale features and capture the long-range dependencies of cracks; (3) a multi-scale features fusion module (MFF) is established to efficiently represent and deeply fuse multi-scale features. We carried out extensive experiments on three pavement crack datasets. These were DeepCrack dataset, with complex background noises; the CrackTree260 dataset, with various crack structures; and the Aerial Track Detection dataset, with a drone's perspective. The experimental results demonstrate that GGMNet has excellent performance, high accuracy, and strong robustness. In conclusion, this paper provides support for accurate and timely road maintenance and has important reference values and enlightening implications for further linear feature extraction research. [ABSTRACT FROM AUTHOR]

UAV transmission tower inspection is the use of UAV technology for regular inspection and troubleshooting of towers on transmission lines, which helps to improve the safety and reliability of transmission lines and ensures the stability of the power supply. From the traditional manual tower boarding to the current way of manually selecting target camera shooting points from 3D point clouds to plan the inspection path of the UAV, operational efficiency has drastically improved. However, indoor planning work is still labor-consuming and expensive. In this paper, a deep learning-based point cloud transmission tower segmentation (PCTTS) model combined with the corresponding target point localization algorithm is proposed for automatic segmentation of transmission tower point cloud data and automatically localizing the key inspection component as the target point for UAV inspection. First, we utilize octree sampling with unit ball normalization to simplify the data and ensure translation invariance before putting the data into the model. In the feature extraction stage, we encode the point set information and combine Euclidean distance and cosine similarity features to ensure rotational invariance. On this basis, we adopt multi-scale feature extraction, construct a local coordinate system, and introduce the offset-attention mechanism to enhance model performance further. Then, after the feature propagation module, gradual up-sampling is used to obtain the features of each point to complete the point cloud segmentation. Finally, combining the segmentation results with the target point localization algorithm completes the automatic extraction of UAV inspection target points. The method has been applied to six kinds of transmission tower point cloud data of part segmentation results and three kinds of transmission tower point cloud data of instance segmentation results. The experimental results show that the model achieves mIOU of 94.1% on the self-built part segmentation dataset and 86.9% on the self-built instance segmentation dataset, and the segmentation accuracy outperforms that of the methods for point cloud segmentation, such as PointNet++, DGCNN, Point Transformer, and PointMLP. Meanwhile, the experimental results of UAV inspection target point localization also verify the method's effectiveness in this paper. [ABSTRACT FROM AUTHOR]

Salient object detection (SOD) aims to identify the most visually striking objects in a scene, simulating the function of the biological visual attention system. The attention mechanism in deep learning is commonly used as an enhancement strategy which enables the neural network to concentrate on the relevant parts when processing input data, effectively improving the model's learning and prediction abilities. Existing saliency object detection methods based on RGB deep learning typically treat all regions equally by using the extracted features, overlooking the fact that different regions have varying contributions to the final predictions. Based on the U2Net algorithm, this paper incorporates the split coordinate channel attention (SCCA) mechanism into the feature extraction stage. SCCA conducts spatial transformation in width and height dimensions to efficiently extract the location information of the target to be detected. While pixel-level semantic segmentation based on annotation has been successful, it assigns the same weight to each pixel which leads to poor performance in detecting the boundary of objects. In this paper, the Canny edge detection loss is incorporated into the loss calculation stage to improve the model's ability to detect object edges. Based on the DUTS and HKU-IS datasets, experiments confirm that the proposed strategies effectively enhance the model's detection performance, resulting in a 0.8% and 0.7% increase in the F1-score of U2Net. This paper also compares the traditional attention modules with the newly proposed attention, and the SCCA attention module achieves a top-three performance in prediction time, mean absolute error (MAE), F1-score, and model size on both experimental datasets. [ABSTRACT FROM AUTHOR]

Plant diseases and pests represent significant threats to agricultural yields, underscoring the need for precise and prompt identification of pathogens and pests. In this paper, we propose a lightweight convolutional neural network model called Gated Asymmetrical GhostNet (GA-GhostNet), specifically designed for the automatic identification of plant diseases and pests. The model incorporates a Gated Multi-scale Coordinate Attention (GM-CA) module to filter out noise and irrelevant information in the image, while capturing location information of diseases and pests at different scales. The model also utilizes an improved feature extraction module called Asymmetrical Ghost (AG) module to enhance robustness to image flipping, as well as improve feature extraction capabilities. Additionally, the CutMix data augmentation method is employed to improve generalization ability. In extensive experimental evaluations on the IP102 pest dataset, GA-GhostNet achieved impressive results with 63.73% MRec, 67.37% MPre, 65.12% MF1, and 71.90% Acc. Moreover, through transfer learning on the Jute pest, Embrapa disease, and Apple disease datasets, the model outperforms numerous lightweight models, attaining accuracies of 99.89%, 96.97%, and 95.17%, respectively, while having only 3.73 million parameters. GA-GhostNet demonstrates high accuracy and efficiency in identifying plant diseases and pests. [ABSTRACT FROM AUTHOR]

Generally, the recognition performance of lightweight models is often lower than that of large models. Knowledge distillation, by teaching a student model using a teacher model, can further enhance the recognition accuracy of lightweight models. In this paper, we approach knowledge distillation from the perspective of intermediate feature-level knowledge distillation. We combine a cross-stage feature fusion symmetric framework, an attention mechanism to enhance the fused features, and a contrastive loss function for teacher and student models at the same stage to comprehensively implement a multistage feature fusion knowledge distillation method. This approach addresses the problem of significant differences in the intermediate feature distributions between teacher and student models, making it difficult to effectively learn implicit knowledge and thus improving the recognition accuracy of the student model. Compared to existing knowledge distillation methods, our method performs at a superior level. On the CIFAR100 dataset, it boosts the recognition accuracy of ResNet20 from 69.06% to 71.34%, and on the TinyImagenet dataset, it increases the recognition accuracy of ResNet18 from 66.54% to 68.03%, demonstrating the effectiveness and generalizability of our approach. Furthermore, there is room for further optimization of the overall distillation structure and feature extraction methods in this approach, which requires further research and exploration. [ABSTRACT FROM AUTHOR]

Ship detection and identification play pivotal roles in ensuring navigation safety and facilitating efficient maritime traffic management. Aiming at ship detection in complex environments, which often faces problems such as the dense occlusion of ship targets, low detection accuracy, and variable environmental conditions, in this paper, we propose a ship detection algorithm CSD-YOLO (Context guided block module, Slim-neck, Deformable large kernel attention-You Only Look Once) based on the deformable large kernel attention (D-LKA) mechanism, which was improved based on YOLOv8 to enhance its performance. This approach integrates several innovations to bolster its performance. Initially, the utilization of the Context Guided Block module (CG block) enhanced the c2f module of the backbone network, thereby augmenting the feature extraction capabilities and enabling a more precise capture of the key image information. Subsequently, the introduction of a novel neck architecture and the incorporation of the slim-neck module facilitated more effective feature fusion, thereby enhancing both the accuracy and efficiency of detection. Furthermore, the algorithm incorporates a D-LKA mechanism to dynamically adjust the convolution kernel shape and size, thereby enhancing the model's adaptability to varying ship target shapes and sizes. To address data scarcity in complex marine environments, the experiments utilized a fused dataset comprising the SeaShips dataset and a proprietary dataset. The experimental results demonstrate that the CSD-YOLO algorithm outperformed the YOLOv8n algorithm across all model evaluation metrics. Specifically, the precision rate (precision) was 91.5%, the recall rate (recall) was 89.5%, and the mean accuracy (mAP) was 91.5%. Compared to the benchmark algorithm, the Recall was improved by 0.7% and the mAP was improved by 0.4%. These results indicate that the CSD-YOLO algorithm can effectively meet the requirements for ship target recognition and tracking in complex marine environments. [ABSTRACT FROM AUTHOR]

Accurate identification of Seriola dumerili (SD) offers crucial technical support for aquaculture practices and behavioral research of this species. However, the task of discerning S. dumerili from complex underwater settings, fluctuating light conditions, and schools of fish presents a challenge. This paper proposes an intelligent recognition model based on the YOLOv8 network called SD-YOLOv8. By adding a small object detection layer and head, our model has a positive impact on the recognition capabilities for both close and distant instances of S. dumerili, significantly improving them. We construct a convenient S. dumerili dataset and introduce the deformable convolution network v2 (DCNv2) to enhance the information extraction process. Additionally, we employ the bottleneck attention module (BAM) and redesign the spatial pyramid pooling fusion (SPPF) for multidimensional feature extraction and fusion. The Inner-MPDIoU bounding box regression function adjusts the scale factor and evaluates geometric ratios to improve box positioning accuracy. The experimental results show that our SD-YOLOv8 model achieves higher accuracy and average precision, increasing from 89.2% to 93.2% and from 92.2% to 95.7%, respectively. Overall, our model enhances detection accuracy, providing a reliable foundation for the accurate detection of fishes. [ABSTRACT FROM AUTHOR]

Traditional fish farming methods suffer from backward production, low efficiency, low yield, and environmental pollution. As a result of thorough research using deep learning technology, the industrial aquaculture model has experienced gradual maturation. A variety of complex factors makes it difficult to extract effective features, which results in less-than-good model performance. This paper proposes a fish detection method that combines a triple attention mechanism with a You Only Look Once (TAM-YOLO)model. In order to enhance the speed of model training, the process of data encapsulation incorporates positive sample matching. An exponential moving average (EMA) is incorporated into the training process to make the model more robust, and coordinate attention (CA) and a convolutional block attention module are integrated into the YOLOv5s backbone to enhance the feature extraction of channels and spatial locations. The extracted feature maps are input to the PANet path aggregation network, and the underlying information is stacked with the feature maps. The method improves the detection accuracy of underwater blurred and distorted fish images. Experimental results show that the proposed TAM-YOLO model outperforms YOLOv3, YOLOv4, YOLOv5s, YOLOv5m, and SSD, with a mAP value of 95.88%, thus providing a new strategy for fish detection. [ABSTRACT FROM AUTHOR]

Traffic detection (including lane detection and traffic sign detection) is one of the key technologies to realize driving assistance system and auto drive system. However, most of the existing detection methods are designed based on single‐modal visible light data, when there are dramatic changes in lighting in the scene (such as insufficient lighting in night), it is difficult for these methods to obtain good detection results. In view of multi‐modal data can provide complementary discriminative information, based on the YoLoV5 model, this paper proposes a multi‐modal fusion YoLoV5 network, which consists of three key components: the dual stream feature extraction module, the correlation feature extraction module, and the self‐attention fusion module. Specifically, the dual stream feature extraction module is used to extract the features of each of the two modalities. Secondly, input the features learned from the dual stream feature extraction module into the correlation feature extraction module to learn the features with maximum correlation. Then, the extracted maximum correlation features are used to achieve information exchange between modalities through a self‐attention mechanism, and thus obtain fused features. Finally, the fused features are inputted into the detection layer to obtain the final detection result. Experimental results on different traffic detection tasks can demonstrate the superiority of the proposed method. [ABSTRACT FROM AUTHOR]

Time series data are prevalent in the real world, particularly playing a crucial role in key domains such as meteorology, electricity, and finance. Comprising observations at historical time points, these data, when subjected to in-depth analysis and modeling, enable researchers to predict future trends and patterns, providing support for decision making. In current research, especially in the analysis of long time series, effectively extracting and integrating long-term dependencies with short-term features remains a significant challenge. Long-term dependencies refer to the correlation between data points spaced far apart in a time series, while short-term features focus on more recent changes. Understanding and combining these two features correctly are crucial for constructing accurate and reliable predictive models. To efficiently extract and integrate long-term dependencies and short-term features in long time series, this paper proposes a pyramid attention structure model based on multi-scale feature extraction, referred to as the MSFformer model. Initially, a coarser-scale construction module is designed to obtain coarse-grained information. A pyramid data structure is constructed through feature convolution, with the bottom layer representing the original data and each subsequent layer containing feature information extracted across different time step lengths. As a result, nodes higher up in the pyramid integrate information from more time points, such as every Monday or the beginning of each month, while nodes lower down retain their individual information. Additionally, a Skip-PAM is introduced, where a node only calculates attention with its neighboring nodes, parent node, and child nodes, effectively reducing the model's time complexity to some extent. Notably, the child nodes refer to nodes selected from the next layer by skipping specific time steps. In this study, we not only propose an innovative time series prediction model but also validate the effectiveness of these methods through a series of comprehensive experiments. To comprehensively evaluate the performance of the designed model, we conducted comparative experiments with baseline models, ablation experiments, and hyperparameter studies. The experimental results demonstrate that the MSFformer model improves by 35.87% and 42.6% on the MAE and MSE indicators, respectively, compared to traditional Transformer models. These results highlight the outstanding performance of our proposed deep learning model in handling complex time series data, particularly in capturing long-term dependencies and integrating short-term features. [ABSTRACT FROM AUTHOR]

In rice cultivation and breeding, obtaining accurate information on the quantity and spatial distribution of rice plants is crucial. However, traditional field sampling methods can only provide rough estimates of the plant count and fail to capture precise plant locations. To address these problems, this paper proposes P2PNet-EFF for the counting and localization of rice plants. Firstly, through the introduction of the enhanced feature fusion (EFF), the model improves its ability to integrate deep semantic information while preserving shallow spatial details. This allows the model to holistically analyze the morphology of plants rather than focusing solely on their central points, substantially reducing errors caused by leaf overlap. Secondly, by integrating efficient multi-scale attention (EMA) into the backbone, the model enhances its feature extraction capabilities and suppresses interference from similar backgrounds. Finally, to evaluate the effectiveness of the P2PNet-EFF method, we introduce the URCAL dataset for rice counting and localization, gathered using UAV. This dataset consists of 365 high-resolution images and 173,352 point annotations. Experimental results on the URCAL demonstrate that the proposed method achieves a 34.87% reduction in MAE and a 28.19% reduction in RMSE compared to the original P2PNet while increasing R 2 by 3.03%. Furthermore, we conducted extensive experiments on three frequently used plant counting datasets. The results demonstrate the excellent performance of the proposed method. [ABSTRACT FROM AUTHOR]

Wearable sensors are widely used in medical applications and human–computer interaction because of their portability and powerful privacy. Human activity identification based on sensor data plays a vital role in these fields. Therefore, it is important to improve the recognition performance of different types of actions. Aiming at the problems of insufficient time-varying feature extraction and gradient explosion caused by too many network layers, a time convolution network recognition model with attention mechanism (TCN-Attention-HAR) was proposed. The model effectively recognizes and emphasizes the key feature information. The ability of extracting temporal features from TCN (temporal convolution network) is improved by using the appropriate size of the receiver domain. In addition, attention mechanisms are used to assign higher weights to important information, enabling models to learn and identify human activities more effectively. The performance of the Open Data Set (WISDM, PAMAP2 and USC-HAD) is improved by 1.13%, 1.83% and 0.51%, respectively, compared with other advanced models, these results clearly show that the network model presented in this paper has excellent recognition performance. In the knowledge distillation experiment, the parameters of student model are only about 0.1% of those of teacher model, and the accuracy of the model has been greatly improved, and in the WISDM data set, compared with the teacher's model, the accuracy is 0.14% higher. [ABSTRACT FROM AUTHOR]

Dynamic monitoring of cropland using high spatial resolution remote sensing images is a powerful means to protect cropland resources. However, when a change detection method based on a convolutional neural network employs a large number of convolution and pooling operations to mine the deep features of cropland, the accumulation of irrelevant features and the loss of key features will lead to poor detection results. To effectively solve this problem, a novel cropland change detection network (CroplandCDNet) is proposed in this paper; this network combines an adaptive receptive field and multiscale feature transmission fusion to achieve accurate detection of cropland change information. CroplandCDNet first effectively extracts the multiscale features of cropland from bitemporal remote sensing images through the feature extraction module and subsequently embeds the receptive field adaptive SK attention (SKA) module to emphasize cropland change. Moreover, the SKA module effectively uses spatial context information for the dynamic adjustment of the convolution kernel size of cropland features at different scales. Finally, multiscale features and difference features are transmitted and fused layer by layer to obtain the content of cropland change. In the experiments, the proposed method is compared with six advanced change detection methods using the cropland change detection dataset (CLCD). The experimental results show that CroplandCDNet achieves the best F1 and OA at 76.04% and 94.47%, respectively. Its precision and recall are second best of all models at 76.46% and 75.63%, respectively. Moreover, a generalization experiment was carried out using the Jilin-1 dataset, which effectively verified the reliability of CroplandCDNet in cropland change detection. [ABSTRACT FROM AUTHOR]

Multi-scale object detection is critical for analyzing remote sensing images. Traditional feature pyramid networks, which are aimed at accommodating objects of varying sizes through multi-level feature extraction, face significant challenges due to the diverse scale variations present in remote sensing images. This situation often forces single-level features to span a broad spectrum of object sizes, complicating accurate localization and classification. To tackle these challenges, this paper proposes an innovative algorithm that incorporates an adaptive multi-scale feature enhancement and fusion module (ASEM), which enhances remote sensing image object detection through sophisticated multi-scale feature fusion. Our method begins by employing a feature pyramid to gather coarse multi-scale features. Subsequently, it integrates a fine-grained feature extraction module at each level, utilizing atrous convolutions with varied dilation rates to refine multi-scale features, which markedly improves the information capture from widely varied object scales. Furthermore, an adaptive enhancement module is applied to the features of each level by employing an attention mechanism for feature fusion. This strategy concentrates on the features of critical scale, which significantly enhance the effectiveness of capturing essential feature information. Compared with the baseline method, namely, Rotated FasterRCNN, our method achieved an mAP of 74.21% (0.81%) on the DOTA-v1.0 dataset and an mAP of 84.90% (+9.2%) on the HRSC2016 dataset. These results validated the effectiveness and practicality of our method and demonstrated its significant application value in multi-scale remote sensing object detection tasks. [ABSTRACT FROM AUTHOR]

The presence of haze results in diminished image clarity, the loss of image details, color alterations, and a decline in image visibility. This paper proposes an efficient multi-stage feature fusion defogging network based on the attention mechanism for effective defogging in complex environments. To create a multi-branch defogging network, the model integrates several attention techniques to create various sorts of branching network architectures. The image's spatial details and contextual information are supplemented and merged based on the recovered image feature information of different network branches to increase the effectiveness of the network model. The stage attention fusion mechanism created among several network branches can lessen the loss of image data during feature extraction and improve the effectiveness of the image-defogging operation. The experimental results demonstrate that the proposed algorithm has superior defogging performance in both synthetic and real-world scene datasets and performs more admirably in terms of accuracy compared to other sophisticated algorithms, particularly in the RESIDE and O-Haze datasets. The PSNR metrics on the RESIDE and O-Haze datasets are improved by 1.58 dB and 1.61 dB, respectively, compared to the best-advanced technique suggested in this study, and the SSIM metrics on the O-Haze dataset are improved by 3.4%. [ABSTRACT FROM AUTHOR]

Domain Name System (DNS) plays an infrastructure role in providing the directory service for mapping domains to IPs on the Internet. Considering the foundation and openness of DNS, it is not surprising that adversaries register massive domains to enable multiple malicious activities, such as spam, command and control (C&C), malware distribution, click fraud, etc. Therefore, detecting malicious domains is a significant topic in security research. Although a substantial quantity of research has been conducted, previous work has failed to fuse multiple relationship features to uncover the deep underlying relationships between domains, thus largely limiting their level of performance. In this paper, we proposed AGCN-Domain to detect malicious domains by combining various relations. The core concept behind our work is to analyze relations between domains according to their behaviors in multiple perspectives and fuse them intelligently. The AGCN-Domain model utilizes three relationships (client relation, resolution relation, and cname relation) to construct three relationship feature graphs to extract features and intelligently fuse the features extracted from the graphs through an attention mechanism. After the relationship features are extracted from the domain names, they are put into the trained classifier to be processed. Through our experiments, we have demonstrated the performance of our proposed AGCN-Domain model. With 10% initialized labels in the dataset, our AGCN-Domain model achieved an accuracy of 94.27% and the F1 score of 87.93%, significantly outperforming other methods in the comparative experiments. [ABSTRACT FROM AUTHOR]

Deep convolutional neural networks (DNNs) have achieved outstanding performance in this field. Meanwhile, handwritten Chinese character recognition (HCCR) is a challenging area of research in the field of computer vision. DNNs require a large number of parameters and high memory consumption. To address these issues, this paper proposes an approach based on an attention mechanism and knowledge distillation. The attention mechanism improves the feature extraction and the knowledge distillation reduces the number of parameters. The experimental results show that ResNet18 achieves a recognition accuracy of 97.63% on the HCCR dataset with 11.25 million parameters. Compared with other methods, this study improves the performance for HCCR. [ABSTRACT FROM AUTHOR]

Head Pose Estimation (HPE) has a wide range of applications in computer vision, but still faces challenges: (1) Existing studies commonly use Euler angles or quaternions as pose labels, which may lead to discontinuity problems. (2) HPE does not effectively address regression via rotated matrices. (3) There is a low recognition rate in complex scenes, high computational requirements, etc. This paper presents an improved unconstrained HPE model to address these challenges. First, a rotation matrix form is introduced to solve the problem of unclear rotation labels. Second, a continuous 6D rotation matrix representation is used for efficient and robust direct regression. The RepVGG-A2 lightweight framework is used for feature extraction, and by adding a multi-level feature fusion module and a coordinate attention mechanism with residual connection, to improve the network's ability to perceive contextual information and pay attention to features. The model's accuracy was further improved by replacing the network activation function and improving the loss function. Experiments on the BIWI dataset 7:3 dividing the training and test sets show that the average absolute error of HPE for the proposed network model is 2.41. Trained on the dataset 300W_LP and tested on the AFLW2000 and BIWI datasets, the average absolute errors of HPE of the proposed network model are 4.34 and 3.93. The experimental results demonstrate that the improved network has better HPE performance. [ABSTRACT FROM AUTHOR]

The joint use of hyperspectral image (HSI) and Light Detection And Ranging (LiDAR) data has been widely applied for land cover classification because it can comprehensively represent the urban structures and land material properties. However, existing methods fail to combine the different image information effectively, which limits the semantic relevance of different data sources. To solve this problem, in this paper, an Attention-guided Fusion and Classification framework based on Convolutional Neural Network (AFC-CNN) is proposed to classify the land cover based on the joint use of HSI and LiDAR data. In the feature extraction module, AFC-CNN employs the three dimensional convolutional neural network (3D-CNN) combined with a multi-scale structure to extract the spatial-spectral features of HSI, and uses a 2D-CNN to extract the spatial features from LiDAR data. Simultaneously, the spectral attention mechanism is adopted to assign weights to the spectral channels, and the cross attention mechanism is introduced to impart significant spatial weights from LiDAR to HSI, which enhance the interaction between HSI and LiDAR data and leverage the fusion information. Then two feature branches are concatenated and transferred to the feature fusion module for higher-level feature extraction and fusion. In the fusion module, AFC-CNN adopts the depth separable convolution connected through the residual structures to obtain the advanced features, which can help reduce computational complexity and improve the fitting ability of the model. Finally, the fused features are sent into the linear classification module for final classification. Experimental results on three datasets, i.e., Houston, MUUFL and Trento datasets show that the proposed AFC-CNN framework achieves better classification accuracy compared with the state-of-the-art algorithms. The overall accuracy of AFC-CNN on Houston, MUUFL and Trento datasets are 94.2%, 95.3% and 99.5%, respectively. [ABSTRACT FROM AUTHOR]

Human activity recognition (HAR) is a crucial detection technique extensively employed in various contexts demanding accurate identification of human actions. Currently, mainstream HAR approaches frequently rely on sequential data generated by wearable sensors, with a primary focus on extracting feature representations that align with various human activities. Nevertheless, complex and diverse human activities encompass various kinds of features and substantial variations in the features emphasized across different activities, making it challenging for current artificial intelligence methods to achieve accurate recognition. To address this challenge, this paper introduces a network framework called the Multi-Feature Combining Attention Neural Network (MFCANN). This framework incorporates diverse feature extraction, along with both local and global feature attention mechanisms. It is composed of stacked Multi-Feature Combining Attention Blocks (MFCAB) designed by us. In contrast to traditional convolutional methods, MFCAB parallelly stacks various convolutional components, enabling the extraction of a broader range of features from human activity data. Additionally, we propose the Intra-Module Attention Block (Intra-MAB) and the Inter-Module Attention Block (Inter-MAB), which simultaneously focus on local fine-grained features within module feature maps and global distinguishing features across module feature maps. This aims to achieve more targeted feature learning, reinforcing the network's ability to distinguish between different human activities. Abundant experimental results demonstrate that the proposed MFCANN in this paper outperforms current mainstream deep learning algorithms in HAR tasks, achieving recognition accuracies of 0.9813, 0.9324, and 0.9930 on the UCI-HAR, USC-HAD, and RealWorld datasets respectively. We release our code at https://github.com/RangerHeart/MFCANN. [ABSTRACT FROM AUTHOR]

Human activity recognition (HAR) is a crucial detection technique extensively employed in various contexts demanding accurate identification of human actions. Currently, mainstream HAR approaches frequently rely on sequential data generated by wearable sensors, with a primary focus on extracting feature representations that align with various human activities. Nevertheless, complex and diverse human activities encompass various kinds of features and substantial variations in the features emphasized across different activities, making it challenging for current artificial intelligence methods to achieve accurate recognition. To address this challenge, this paper introduces a network framework called the Multi-Feature Combining Attention Neural Network (MFCANN). This framework incorporates diverse feature extraction, along with both local and global feature attention mechanisms. It is composed of stacked Multi-Feature Combining Attention Blocks (MFCAB) designed by us. In contrast to traditional convolutional methods, MFCAB parallelly stacks various convolutional components, enabling the extraction of a broader range of features from human activity data. Additionally, we propose the Intra-Module Attention Block (Intra-MAB) and the Inter-Module Attention Block (Inter-MAB), which simultaneously focus on local fine-grained features within module feature maps and global distinguishing features across module feature maps. This aims to achieve more targeted feature learning, reinforcing the network's ability to distinguish between different human activities. Abundant experimental results demonstrate that the proposed MFCANN in this paper outperforms current mainstream deep learning algorithms in HAR tasks, achieving recognition accuracies of 0.9813, 0.9324, and 0.9930 on the UCI-HAR, USC-HAD, and RealWorld datasets respectively. We release our code at https://github.com/RangerHeart/MFCANN. [ABSTRACT FROM AUTHOR]

The pivotal factor to regulate and enhance within an operating batch is the quality process, a challenging variable to monitor online. Soft sensors offer an immediate alternative for providing real-time insights into process quality, yet persisting issues include the imbalance between process and quality measurements, noisy measurements, and the intricate 3-dimensional dynamic batch data structure. This paper introduces the Robust Semi-Supervised Dual-Attention Latent Dynamic Conditional State-Space Model (RS2DA-LDCSSM) to address these challenges for within-batch quality prediction. Given the frequent absence of quality data due to measurement inconveniences, an imputation network is embedded within the RS2DA-LDCSSM to facilitate the prediction of future quality. To prevent information distortion during data unfolding, the Attentional Sequence-to-Sequence RNN Encoder-Decoder (AS2S-RNNED) is employed to process the 3-dimensional batch data. The proposed method integrates AS2S-RNNED with a probability state-space model to filter out noisy process data and stabilize the probability prediction of quality data, extracting spatial and temporal latent data from past process and quality data while minimizing the loss in multi-step prediction. This work represents a novel approach to dynamic nonlinear batch processes through probabilistic semi-supervised learning. The RS2DA-LDCSSM is adaptable to any batch process for within-batch quality prediction, as evidenced by numerical cases demonstrating its reliability, achieving an R2 value of 0.87, surpassing comparison methods. In an industrial penicillin fermentation batch process, the RS2DA-LDCSSM exhibits remarkable robustness with a high R2 index value of 0.99 in practical quality prediction scenarios, showcasing its efficacy. • Semi-supervised model training utilizes unlabeled process data. • Separate networks are crafted for process and quality feature extraction. • The state-space model with the encoder-decoder is used to extract latent features. • A multi-step-ahead prediction learning scheme yields favorable prediction results. [ABSTRACT FROM AUTHOR]

The pivotal factor to regulate and enhance within an operating batch is the quality process, a challenging variable to monitor online. Soft sensors offer an immediate alternative for providing real-time insights into process quality, yet persisting issues include the imbalance between process and quality measurements, noisy measurements, and the intricate 3-dimensional dynamic batch data structure. This paper introduces the Robust Semi-Supervised Dual-Attention Latent Dynamic Conditional State-Space Model (RS2DA-LDCSSM) to address these challenges for within-batch quality prediction. Given the frequent absence of quality data due to measurement inconveniences, an imputation network is embedded within the RS2DA-LDCSSM to facilitate the prediction of future quality. To prevent information distortion during data unfolding, the Attentional Sequence-to-Sequence RNN Encoder-Decoder (AS2S-RNNED) is employed to process the 3-dimensional batch data. The proposed method integrates AS2S-RNNED with a probability state-space model to filter out noisy process data and stabilize the probability prediction of quality data, extracting spatial and temporal latent data from past process and quality data while minimizing the loss in multi-step prediction. This work represents a novel approach to dynamic nonlinear batch processes through probabilistic semi-supervised learning. The RS2DA-LDCSSM is adaptable to any batch process for within-batch quality prediction, as evidenced by numerical cases demonstrating its reliability, achieving an R2 value of 0.87, surpassing comparison methods. In an industrial penicillin fermentation batch process, the RS2DA-LDCSSM exhibits remarkable robustness with a high R2 index value of 0.99 in practical quality prediction scenarios, showcasing its efficacy. • Semi-supervised model training utilizes unlabeled process data. • Separate networks are crafted for process and quality feature extraction. • The state-space model with the encoder-decoder is used to extract latent features. • A multi-step-ahead prediction learning scheme yields favorable prediction results. [ABSTRACT FROM AUTHOR]

Endoscopic ultrasonography (EUS) is a valuable imaging modality for diagnosing gastrointestinal submucosal tumors (SMTs). However, inherent content variations in EUS images due to gastrointestinal tract mobility and handheld ultrasound instability challenge the saliency of SMTs' visual features. The presence of fine-grained inter-class and large intra-class differences further complicates EUS-based diagnosis. To address these issues, this paper presents a novel Multi-Attribute Guided Contextual Attention Network (MAG-CA-Net) for interpretable SMT recognition in EUS. Inspired by endoscopists' clinical diagnosis expertise, our framework initially localizes abnormal areas based on echo attributes and subsequently determines tumor categories using contextual semantics. Experimental results demonstrate the effectiveness of MAG-CA-Net, exhibiting improved recognition recall and precision rates for gastrointestinal stromal tumor, leiomyoma, and pancreatic rest. Specifically, the MAG network facilitates abnormal area localization, while the CA mechanism guides the model to focus on the most discriminative tumor-context-associated regions. The proposed method achieved an average classification accuracy of 93.16%, an average precision of 93.17%, a weighted recall of 93.16%, and an average F1-score of 93.15 % for the three disease categories. The proposed approach provides crucial guidelines for data collection standards and model development in the clinical diagnosis process of SMTs under EUS. Its interpretability analysis enhances the credibility of clinical physicians towards assisted diagnostic methods based on deep learning. The source code will be publicly available at https://gitee.com/HangbinZheng/mag-ca-net. • Hierarchical identification strategy combining clinical knowledge and data-driven models for submucosal tumor detection. • Contextual attention mechanism enhances tumor recognition using gastrointestinal layered features. • Improved adaptability to endoscopic ultrasound image variation and fine-grained feature extraction compared to state-of-the-art models. • Visual interpretability close to experienced endoscopists enhances method's interpretability. [ABSTRACT FROM AUTHOR]

In this paper, we propose a 3D object detection method called MMAF-Net that is based on the multi-view and multi-stage adaptive fusion of RGB images and LiDAR point cloud data. This is an end-to-end architecture, which combines the characteristics of RGB images, the front view of point clouds based on reflection intensity, and the bird's eye view of point clouds. It also adopts a multi-stage fusion approach of "data-level fusion + feature-level fusion" to fully exploit the strength of multimodal information. Our proposed method addresses key challenges found in current 3D object detection methods for autonomous driving, including insufficient feature extraction from multimodal data, rudimentary fusion techniques, and sensitivity to distance and occlusion. To ensure the comprehensive integration of multimodal information, we present a series of targeted fusion methods. Firstly, we propose a novel input form that encodes dense point cloud reflectivity information into the image to enhance its representational power. Secondly, we design the Region Attention Adaptive Fusion module utilizing an attention mechanism to guide the network in adaptively adjusting the importance of different features. Finally, we extend the 2D DIOU (Distance Intersection over Union) loss function to 3D and develop a joint regression loss based on 3D_DIOU and SmoothL1 to optimize the similarity between detected and ground truth boxes. The experimental results on the KITTI dataset demonstrate that MMAF-Net effectively addresses the challenges posed by highly obscured or crowded scenes while maintaining real-time performance and improving the detection accuracy of smaller and more difficult objects that are occluded at far distances. • A novel object detection method is proposed integrating RGB images and LiDAR data. • This study tests the method on the KITTI dataset. • This method tackles highly obscured or crowded scenes for autonomous driving. • This method improves the detection accuracy of smaller and occluded targets far away. [ABSTRACT FROM AUTHOR]

Remote sensing scene understanding is crucial for extracting valuable information from high-resolution images, including object detection and classification. Traditional object detection methods face challenges in handling the diverse scales, orientations, and complex backgrounds present in remote sensing data. In this paper, we propose a novel remote sensing scene understanding system called multiscale-attention R-CNN (MSA R-CNN), which incorporates a super multiscale feature extraction network (SMENet) for enhanced feature extraction from multiscale images, an adaptive dynamic inner lateral (ADIL) connection module to tackle information loss in feature pyramid networks (FPN), and a distributed lightweight attention module (DLAM) to refine feature information processing. Furthermore, a new dataset combining the DIOR and DOTA datasets is introduced to extract the background information of detected objects and evaluate the proposed system's performance. MSA R-CNN achieved an mAP of 74.37% on the DIOR dataset when the gamma value was set to 0.2 and 81.97% on the DOTA dataset when the gamma value was set to 0.1 with the same learning rate, outperforming state-of-the-art models on both datasets. The proposed system demonstrates significant improvements in both object detection and background information extraction, providing a comprehensive solution for remote sensing scene understanding. • MSA R-CNN is proposed for remote sensing scene understanding. • SMENet enhances multi-scale feature extraction in MSA R-CNN. • ADIL module tackles FPN information loss. • DLAM refines feature information processing in FPN. • Achieved state-of-the-art performance on DIOR & DOTA dataset. [ABSTRACT FROM AUTHOR]

In integrated energy systems (IESs), reliable planning and operation are challenging owing to significant uncertainties in energy production, utilization, and trading. To this end, this paper proposes a multi-task joint forecasting method that enables joint source-load-price forecasting. First, three uncertain variables in an IES, namely, renewable energy, the multi-energy load, and the energy price, were investigated and the complex coupling relationships among them were validated. Second, to cope with the redundant noise resulting from various inputs, multi-channel feature extraction and a hybrid attention mechanism were combined to enable separate extraction and unified fusion of features. Additionally, considering the unique one-dimensional input in the prediction domain, a sequential convolution attention module (SCAM) with a hybrid channel and temporal attention mechanism was proposed to guide multi-channel feature fusion. Finally, facing the challenge of multi-layer coupling information learning, a multi-task learning (MTL) integrated shared layer was designed. Based on the coordinated with MTL, multi-column convolutional neural network, SCAM and long short-term memory network, joint forecasting of source-load-price was realized. The simulation results showed that the average mean absolute percentage error of the proposed model was as low as 4.10% in source-load-price long-term forecasting, while that of winter short-term forecasting could reach 3.14%. In addition, the here proposed model was found to be superior to others in terms of computational efficiency and result stability. • Combining energy and market layers for source-load-price joint forecasting. • Proposing a SCAM with hybrid channel and temporal attention modules. • Designing an MTL integrated sharing layer and achieving adaptive adjusting of the loss function. • The average MAPE of the proposed model is as low as 4.10% for the long-term source-load-price forecasting. [ABSTRACT FROM AUTHOR]

Medical imaging is an important tool for clinical diagnosis. Nevertheless, it is very time-consuming and error-prone for physicians to prepare imaging diagnosis reports. Therefore, it is necessary to develop some methods to generate medical imaging reports automatically. Currently, the task of medical imaging report generation is challenging in at least two aspects: (1) medical images are very similar to each other. The differences between normal and abnormal images and between different abnormal images are usually trivial; (2) unrelated or incorrect keywords describing abnormal findings in the generated reports lead to mis-communications. In this paper, we propose a medical image report generation framework composed of four modules, including a Transformer encoder, a MIX-MLP multi-label classification network, a co-attention mechanism (CAM) based semantic and visual feature fusion, and a hierarchical LSTM decoder. The Transformer encoder can be used to learn long-range dependencies between images and labels, effectively extract visual and semantic features of images, and establish long-term dependent relationships between visual and semantic information to accurately extract abnormal features from images. The MIX-MLP multi-label classification network, the co-attention mechanism and the hierarchical LSTM network can better identify abnormalities, achieving visual and text alignment fusion and multi-label diagnostic classification to better facilitate report generation. The results of the experiments performed on two widely used radiology report datasets, IU X-RAY and MIMIC-CXR, show that our proposed framework outperforms current report generation models in terms of both natural linguistic generation metrics and clinical efficacy assessment metrics. The code of this work is available online at https://github.com/watersunhznu/LIFMRG. • A label information fused medical image report generation framework. • ViT encoder based low-level feature extraction for medical image report generation. • A MIX-MLP based multi-label classification network for medical image tag classification. • A POS-SCAN based co-attention mechanism that aligns and fuses both visual information and semantic information. • The focal loss function for the multi-label classification network to address the problem of imbalanced and sparse label distribution. [ABSTRACT FROM AUTHOR]

With the rapid advancements in computer vision object detection networks, several detection networks based on remote sensing data have been proposed for oriented object detection. However, existing oriented networks have limitations in terms of feature extraction and oriented information extraction. To address these limitations, in this paper, we proposed an advanced adaptive points-based anchor-free oriented detection network called RA 2 DC-Net which includes a residual augment-convolutions (Res-AugConvs) module that uses residual structures and attention mechanisms to enhance feature focus information and minimize the loss of focus feature information. In addition, we proposed a method called adaptive deformable convolution (ADConv) to fine-tune custom convolution weights according to the features and generate high-quality adaptive offsets for accurate orientation information. Furthermore, we evaluated the performance of RA 2 DC-Net on two large-scale datasets: the Dataset for Object deTection in Aerial images (DOTA), the rotated object DetectIon in Optical Remote sensing images (DIOR-R) and the High-Resolution Ship Collection 2016 (HRSC2016). Experimental results demonstrated that RA 2 DC-Net achieved mAP values of 76.25%, 68.73% and 89.75% on the three datasets, respectively, with high AP values for various object classes on DOTA. • Decreased the continuous convolutions losses of oriented information in oriented object detection. • Enabling adaptive orientations of object features improves the precision of offset calculations. • Devising a new head structure facilitates focused oriented object features. [ABSTRACT FROM AUTHOR]

Induction motors are widely used in various industrial applications due to their simplicity, robustness, and high efficiency. In recent years, deep learning approaches have shown great potential for detecting and diagnosing faults in induction motors. However, it was observed that the convolutional neural network (CNN) might miss the crucial long-range temporal dependencies of the significant features. At the same time, the recursive neural network (RNN) may face issues such as vanishing gradient and less computational efficiency if the data is too long with complex patterns. This paper proposes a novel deep-learning model that combines a one-dimensional CNN and an RNN into two independent pipelines. The spatial, temporal, long-term, and short-term features can be extracted from the raw input signal. A multi-head mechanism, a state-of-the-art deep learning technique in natural language processing (NLP), enables the model to focus more on relevant features and prevent overfitting. Specifically, our method simplifies the conventional fault diagnosis workflow while maintaining competitive accuracy compared to several state-of-the-art methods. By leveraging the ability to detect faults directly from raw, unprocessed electrical signals, our approach offers a practical, cost-effective solution for real-time fault detection in industrial applications. This innovation has the potential to significantly reduce downtime and the associated operational costs, thereby providing immediate practical value in various industrial settings. [ABSTRACT FROM AUTHOR]

Wind power serves as a clean and sustainable form of energy. However, its generation is fraught with variability and uncertainty, owing to the stochastic and dynamic characteristics of wind. Accurate forecasting of wind power is indispensable for the efficient planning, operation, and grid integration of wind energy systems. In this paper, we introduce a novel forecasting method termed Temporal Collaborative Attention (TCOAT). This data-driven approach is designed to capture both temporal and spatial dependencies in wind power generation data, as well as discern long-term and short-term patterns. Utilizing attention mechanisms, TCOAT dynamically adjusts the weights of each input variable and time step based on their contextual relevance for forecasting. Furthermore, the method employs collaborative attention units to assimilate directional and global information from the input data. It also explicitly models the interactions and correlations among different variables or time steps through the use of self-attention and cross-attention mechanisms. To integrate long-term and short-term information effectively, TCOAT incorporates a temporal fusion layer that employs concatenation and mapping operations, along with hierarchical feature extraction and aggregation. We validate the efficacy of TCOAT through extensive experiments on a real-world wind power generation dataset from Greece and compare its performance against twenty-two state-of-the-art methods. Experimental results demonstrate that TCOAT outperforms existing methods in terms of both accuracy and robustness in wind power forecasting. Moreover, we conduct a generality study on an additional real-world dataset from a different climate condition and wind power characteristics. The results show that TCOAT can achieve comparable or better performance than the state-of-the-art methods, confirming the generalization ability of TCOAT. • Novel deep learning model for wind power forecasting • Temporal collaborative attention and fusion for wind data • Superior performance on real-world wind power dataset • Attention maps for interpretability and end-to-end training [ABSTRACT FROM AUTHOR]

Skin cancer is the most prevalent type of cancer worldwide. Early detection is essential as it could be fatal at later stages. The classification of skin lesions is challenging since there are many variations, including changes in color, shape, size, high intra-class variation, and high inter-class similarity. In this paper, a unique class-wise attention method is proposed that considers each class equally while extracting additional discriminative information of skin lesions. The proposed attention mechanism is employed in a progressive manner to incorporate discriminative feature information from multiple scales. The proposed approach obtained competitive performance against more than 15 state-of-the-art methods including HAM1000 and ISIC 2019 leaderboard winners. The proposed method achieved 97.40% accuracy on the HAM10000 and 94.9% accuracy on the ISIC 2019 dataset. • A novel attention based mechanism is proposed for skin lesions classification. • The problem of significant intra-class variance, high inter-class similarity, and class imbalance is addressed by a class-wise attention mechanism. • The important features for the classification are extracted without the burden of additional parameters using a progressive class-wise attention mechanism. • The final three layers of the baseline model are discarded and the global average pooling (GAP) and classification output layers are added to reduce the number of parameters, while maintaining performance. • The proposed network is robust, end-to-end trainable and has good generalization on unseen data. [ABSTRACT FROM AUTHOR]

Automatic extraction of flame area plays an important role in forest fire detection, which can accurately understand the spatial distribution and development trend of forest fire, so as to effectively realize the protection of forest resources. However, due to the instability and spread of fires, and the complexity of the background, accurate early fire detection is extremely challenging. At the same time, the image pixel proportion of the flame area in early stage is much smaller than that in the background, which causes a serious class imbalance problem. With the fast development of deep learning, some achievements have been made in flame extraction, but there are still some deficiencies in the existing networks, such as limited feature representation, poor feature capturing ability on micro objects, insufficiency processing of local features, etc. This paper proposes an attention-based dual-encoding segmentation network, abbreviated as ADE-Net, for pixelwise early fire detection. To realize strong feature representation, a dual-encoding path, consisting of semantic units and spatial units, is introduced to extract richer features, and an attention fusion module (AFM) is introduced to fully integrate spatial and semantic information and achieve effective feature aggregation. In addition, faced with the class imbalance problem, a multi-attention fusion (MAF) module is introduced to obtain more discriminating features to make the segmentation network to focus on the key pixel areas. Furthermore, a feature enhancement module, named attention-guided enhancement (AGE) module, is proposed to enrich the feature representation of local feature maps. Finally, to realize better multi-scale global feature extraction and fusion, a global context fusion (GCF) module is proposed into the bottleneck layer for multi-scale feature enhancement. Experimental results show that the proposed ADE-Net has a good early fire detection ability from remote sensing images, and it has obtained a competitive advantage compared with advanced segmentation models. [ABSTRACT FROM AUTHOR]

Automatic pollen detection based on light microscope (LM) images is helpful for pollinosis symptoms prevention. Recently, many deep learning methods have been proposed to identify pollen grains based on multi-scale feature fusion mechanism. However, in real scenarios, there are two main challenges that need to be considered: (1) Complex pollen characteristics; (2) Irrelevant objects interference. It means that the pollen detection requires not only learning the relationship among multi-scale features but also refining the feature representation. To this end, this paper proposes an attention-based multi-scale feature fusion network (AMFF-Net) for automatic pollen detection on real-world LM images. The proposed AMFF-Net includes three modules: The feature extraction module utilizes the series-connection attention to capture the spatial and channel dependencies of different level feature maps (for solving the challenge 1). In the feature fusion module, a parallel connection attention is able to learn more discriminative feature representation based on bidirectional pathway guidance (for solving the challenge 2). Both are jointly adopted to enhance the representational capacity of the final results in the pollen prediction module. Extensive experiments are conducted on the real-world RPD dataset, and our AMFF-Net achieves the best performance (83.9% of mean average precision) comparing with other state-of-the-art methods. We believe that this work can serve as an important reference for the development of pollen monitoring system in a real scenario. • An attention-based feature fusion network is proposed for pollen detection. • Series connection attention aims to learn multiple dependencies of feature maps. • Parallel connection attention is used to extract discriminative features. • Both are jointly adopted to enhance the representational capacity. • Extensive Experiments are conducted to prove the effectiveness of our method. [ABSTRACT FROM AUTHOR]