Search

The detection algorithm commonly misses obscured pedestrians in traffic scenes with a high pedestrian density because mutual occlusion among pedestrians reduces the prediction box score of the concealed pedestrians. The paper uses the YOLOv7 algorithm as the baseline and makes the following three improvements by investigating the variables influencing the detection method's performance: First, the backbone network of the YOLOv7 algorithm is replaced with the lightweight feature extraction network Mobilenetv3 since the pedestrian detection algorithm frequently needs to be deployed in driverless mobile, which requires a fast operating speed of the algorithm; second, a high-resolution feature pyramid structure is suggested for the issue of missed detection of hidden pedestrians, which upscales the feature maps generated from the feature pyramid to increase the resolution of the output feature maps and introduces shallow feature maps to strengthen the distinctions between adjacent sub-features to enhance the network's ability to extract features for the visible area of hidden pedestrians and small-sized pedestrians in order to produce deeper features with greater differentiation for pedestrians; and the third is to suggest a detection head based on an attention mechanism that is employed to lower the confidence level of target neighboring sub-features, lower the quantity of redundant detection boxes, and lower the following NMS computation. The mAP of the suggested approach in this work achieves 89.75%, which is 9.5 percentage points better than the YOLOv7 detection algorithm, according to experiments on the CrowdHuman pedestrian-intensive dataset. The algorithm proposed in this paper can considerably increase the detection performance of the detection algorithm, particularly for obscured pedestrians and small-sized pedestrians in the dataset, according to the experimental effect plots. [ABSTRACT FROM AUTHOR]

This paper proposes a multimodal fusion 3D target detection algorithm based on the attention mechanism to improve the performance of 3D target detection. The algorithm utilizes point cloud data and information from the camera. For image feature extraction, the ResNet50 + FPN architecture extracts features at four levels. Point cloud feature extraction employs the voxel method and FCN to extract point and voxel features. The fusion of image and point cloud features is achieved through regional point fusion and voxel fusion methods. After information fusion, the Coordinate and SimAM attention mechanisms extract fusion features at a deep level. The algorithm's performance is evaluated using the DAIR-V2X dataset. The results show that compared to the Part-A2 algorithm; the proposed algorithm improves the mAP value by 7.9% in the BEV view and 7.8% in the 3D view at IOU = 0.5 (cars) and IOU = 0.25 (pedestrians and cyclists). At IOU = 0.7 (cars) and IOU = 0.5 (pedestrians and cyclists), the mAP value of the SECOND algorithm is improved by 5.4% in the BEV view and 4.3% in the 3D view, compared to other comparison algorithms. [ABSTRACT FROM AUTHOR]

With the advancement of artificial intelligence technology and computer hardware, the stereo matching algorithm has been widely researched and applied in the field of image processing. In scenarios such as robot navigation and autonomous driving, stereo matching algorithms are used to assist robots in acquiring depth information about the surrounding environment, thereby improving the robot's ability for autonomous navigation during self-driving. In this paper, we address the issue of low matching accuracy of stereo matching algorithms in specular regions of images and propose a multi-attention-based stereo matching algorithm called MANet. The proposed algorithm embeds a multi-spectral attention module into the residual feature-extraction network of the PSMNet algorithm. It utilizes different 2D discrete cosine transforms to extract frequency-specific feature information, providing rich and effective features for cost computation in matching. The pyramid pooling module incorporates a coordinated attention mechanism, which not only maintains long-range dependencies with directional awareness but also captures more positional information during the pooling process, thereby enhancing the network's representational capacity. The MANet algorithm was evaluated on three major benchmark datasets, namely, SceneFlow, KITTI2015, and KITTI2012, and compared with relevant algorithms. Experimental results demonstrated that the MANet algorithm achieved higher accuracy in predicting disparities and exhibited stronger robustness against specular reflections, enabling more accurate disparity prediction in specular regions. [ABSTRACT FROM AUTHOR]

The detection and classification of bone marrow (BM) cells is a critical cornerstone for hematology diagnosis. However, the low accuracy caused by few BM-cell data samples, subtle difference between classes, and small target size, pathologists still need to perform thousands of manual identifications daily. To address the above issues, we propose an improved BM-cell-detection algorithm in this paper, called YOLOv7-CTA. Firstly, to enhance the model's sensitivity to fine-grained features, we design a new module called CoTLAN in the backbone network to enable the model to perform long-term modeling between target feature information. Then, in order to cooperate with the CoTLAN module to pay more attention to the features in the area to be detected, we integrate the coordinate attention (CoordAtt) module between the CoTLAN modules to improve the model's attention to small target features. Finally, we cluster the target boxes of the BM cell dataset based on K-means++ to generate more suitable anchor boxes, which accelerates the convergence of the improved model. In addition, in order to solve the imbalance between positive and negative samples in BM-cell pictures, we use the Focal loss function to replace the multi-class cross entropy. Experimental results demonstrate that the best mean average precision (mAP) of the proposed model reaches 88.6%, which is an improvement of 12.9%, 8.3%, and 6.7% compared with that of the Faster R-CNN model, YOLOv5l model, and YOLOv7 model, respectively. This verifies the effectiveness and superiority of the YOLOv7-CTA model in BM-cell-detection tasks. [ABSTRACT FROM AUTHOR]

The application of mulching film has significantly contributed to improving agricultural output and benefits, but residual film has caused severe impacts on agricultural production and the environment. In order to realize the accurate recycling of agricultural residual film, the detection of residual film is the first problem to be solved. The difference in color and texture between residual film and bare soil is not obvious, and residual film is of various sizes and morphologies. To solve these problems, the paper proposes a method for detecting residual film in agricultural fields that uses the attention mechanism. First, a two-stage pre-training approach with strengthened memory is proposed to enable the model to better understand the residual film features with limited data. Second, a multi-scale feature fusion module with adaptive weights is proposed to enhance the recognition of small targets of residual film by using attention. Finally, an inter-feature cross-attention mechanism that can realize full interaction between shallow and deep feature information to reduce the useless noise extracted from residual film images is designed. The experimental results on a self-made residual film dataset show that the improved model improves precision, recall, and mAP by 5.39%, 2.02%, and 3.95%, respectively, compared with the original model, and it also outperforms other recent detection models. The method provides strong technical support for accurately identifying farmland residual film and has the potential to be applied to mechanical equipment for the recycling of residual film. [ABSTRACT FROM AUTHOR]

Most deep-learning-based object detection algorithms exhibit low speeds and accuracy in gear surface defect detection due to their high computational costs and complex structures. To solve this problem, a lightweight model for gear surface defect detection, namely STMS-YOLOv5, is proposed in this paper. Firstly, the ShuffleNetv2 module is employed as the backbone to reduce the giga floating-point operations per second and the number of parameters. Secondly, transposed convolution upsampling is used to enhance the learning capability of the network. Thirdly, the max efficient channel attention mechanism is embedded in the neck to compensate for the accuracy loss caused by the lightweight backbone. Finally, the SIOU_Loss is adopted as the bounding box regression loss function in the prediction part to speed up the model convergence. Experiments show that STMS-YOLOv5 achieves frames per second of 130.4 and 133.5 on the gear and NEU-DET steel surface defect datasets, respectively. The number of parameters and GFLOPs are reduced by 44.4% and 50.31%, respectively, while the mAP@0.5 reaches 98.6% and 73.5%, respectively. Extensive ablation and comparative experiments validate the effectiveness and generalization capability of the model in industrial defect detection. [ABSTRACT FROM AUTHOR]

Since light propagation in water bodies is subject to absorption and scattering effects, underwater images using only conventional intensity cameras will suffer from low brightness, blurred images, and loss of details. In this paper, a deep fusion network is applied to underwater polarization images; that is, the underwater polarization images are fused with intensity images using the deep learning method. To construct a training dataset, we establish an experimental setup to obtain underwater polarization images and perform appropriate transformations to expand the dataset. Next, an end-to-end learning framework based on unsupervised learning and guided by an attention mechanism is constructed for fusing polarization and light intensity images. The loss function and weight parameters are elaborated. The produced dataset is used to train the network under different loss weight parameters, and the fused images are evaluated based on different image evaluation metrics. The results show that the fused underwater images are more detailed. Compared with light intensity images, the information entropy and standard deviation of the proposed method increase by 24.48% and 139%. The image processing results are better than other fusion-based methods. In addition, the improved U-net network structure is used to extract features for image segmentation. The results show that the target segmentation based on the proposed method is feasible under turbid water. The proposed method does not require manual adjustment of weight parameters, has faster operation speed, and has strong robustness and self-adaptability, which is important for research in vision fields, such as ocean detection and underwater target recognition. [ABSTRACT FROM AUTHOR]

Safety helmet wearing plays a major role in protecting the safety of workers in industry and construction, so a real-time helmet wearing detection technology is very necessary. This paper proposes an improved YOLOv4 algorithm to achieve real-time and efficient safety helmet wearing detection. The improved YOLOv4 algorithm adopts a lightweight network PP-LCNet as the backbone network and uses deepwise separable convolution to decrease the model parameters. Besides, the coordinate attention mechanism module is embedded in the three output feature layers of the backbone network to enhance the feature information, and an improved feature fusion structure is designed to fuse the target information. In terms of the loss function, we use a new SIoU loss function that fuses directional information to increase detection precision. The experimental findings demonstrate that the improved YOLOv4 algorithm achieves an accuracy of 92.98%, a model size of 41.88 M, and a detection speed of 43.23 pictures/s. Compared with the original YOLOv4, the accuracy increases by 0.52%, the model size decreases by about 83%, and the detection speed increases by 88%. Compared with other existing methods, it performs better in terms of precision and speed. [ABSTRACT FROM AUTHOR]

The accuracy of insulators and their defect identification by UAVs (unmanned aerial vehicles) in transmission-line inspection needs to be further improved, and the model size of the detection algorithm is significantly reduced to make it more suitable for edge-end deployment. In this paper, the algorithm uses a lightweight GhostNet module to reconstruct the backbone feature extraction network of the YOLOv4 model and employs depthwise separable convolution in the feature fusion layer. The model is lighter on the premise of ensuring the effect of image information extraction. Meanwhile, the ECA-Net channel attention mechanism is embedded into the feature extraction layer and PANet (Path Aggregation Network) to improve the recognition accuracy of the model for small targets. The experimental results show that the size of the improved model is reduced from 244 MB to 42 MB, which is only 17.3% of the original model. At the same time, the mAp of the improved model is 0.77% higher than that of the original model, reaching 95.4%. Moreover, the mAP compared with YOLOv5-s and YOLOX-s, respectively, is improved by 1.98% and 1.29%. Finally, the improved model is deployed into Jetson Xavier NX and run at a speed of 8.8 FPS, which is 4.3 FPS faster than the original model. [ABSTRACT FROM AUTHOR]

Light field (LF) image depth estimation is a critical technique for LF-related applications such as 3D reconstruction, target detection, and tracking. The refocusing property of LF images provide rich information for depth estimations; however, it is still challenging in cases of occlusion regions, edge regions, noise interference, etc. The epipolar plane image (EPI) of LF can effectively deal with the depth estimation because of its characteristics of multidirectionality and pixel consistency—in which the LF depth estimations are converted to calculate the EPI slope. This paper proposed an EPI LF depth estimation algorithm based on a directional relationship model and attention mechanism. Unlike the subaperture LF depth estimation method, the proposed method takes EPIs as input images. Specifically, a directional relationship model was used to extract direction features of the horizontal and vertical EPIs, respectively. Then, a multiviewpoint attention mechanism combining channel attention and spatial attention is used to give more weight to the EPI slope information. Subsequently, multiple residual modules are used to eliminate the redundant features that interfere with the EPI slope information—in which a small stride convolution operation is used to avoid losing key EPI slope information. The experimental results revealed that the proposed algorithm outperformed the compared algorithms in terms of accuracy. [ABSTRACT FROM AUTHOR]

A recommendation system is often used to recommend items that may be of interest to users. One of the main challenges is that the scarcity of actual interaction data between users and items restricts the performance of recommendation systems. To solve this problem, multi-modal technologies have been used for expanding available information. However, the existing multi-modal recommendation algorithms all extract the feature of single modality and simply splice the features of different modalities to predict the recommendation results. This fusion method can not completely mine the relevance of multi-modal features and lose the relationship between different modalities, which affects the prediction results. In this paper, we propose a Cross-Modal-Based Fusion Recommendation Algorithm (CMBF) that can capture both the single-modal features and the cross-modal features. Our algorithm uses a novel cross-modal fusion method to fuse the multi-modal features completely and learn the cross information between different modalities. We evaluate our algorithm on two datasets, MovieLens and Amazon. Experiments show that our method has achieved the best performance compared to other recommendation algorithms. We also design ablation study to prove that our cross-modal fusion method improves the prediction results. [ABSTRACT FROM AUTHOR]

In this paper, we propose a multi-channel cross-tower with attention mechanisms in latent domain network (Multi-TALK) that suppresses both the acoustic echo and background noise. The proposed approach consists of the cross-tower network, a parallel encoder with an auxiliary encoder, and a decoder. For the multi-channel processing, a parallel encoder is used to extract latent features of each microphone, and the latent features including the spatial information are compressed by a 1D convolution operation. In addition, the latent features of the far-end are extracted by the auxiliary encoder, and they are effectively provided to the cross-tower network by using the attention mechanism. The cross tower network iteratively estimates the latent features of acoustic echo and background noise in each tower. To improve the performance at each iteration, the outputs of each tower are transmitted as the input for the next iteration of the neighboring tower. Before passing through the decoder, to estimate the near-end speech, attention mechanisms are further applied to remove the estimated acoustic echo and background noise from the compressed mixture to prevent speech distortion by over-suppression. Compared to the conventional algorithms, the proposed algorithm effectively suppresses the acoustic echo and background noise and significantly lowers the speech distortion. [ABSTRACT FROM AUTHOR]

In recent years, increasing image data comes from various sensors, and object detection plays a vital role in image understanding. For object detection in complex scenes, more detailed information in the image should be obtained to improve the accuracy of detection task. In this paper, we propose an object detection algorithm by jointing semantic segmentation (SSOD) for images. First, we construct a feature extraction network that integrates the hourglass structure network with the attention mechanism layer to extract and fuse multi-scale features to generate high-level features with rich semantic information. Second, the semantic segmentation task is used as an auxiliary task to allow the algorithm to perform multi-task learning. Finally, multi-scale features are used to predict the location and category of the object. The experimental results show that our algorithm substantially enhances object detection performance and consistently outperforms other three comparison algorithms, and the detection speed can reach real-time, which can be used for real-time detection. [ABSTRACT FROM AUTHOR]

Passenger flow prediction has drawn increasing attention in the deep learning research field due to its great importance in traffic management and public safety. The major challenge of this essential task lies in multiple spatiotemporal correlations that exhibit complex non-linear correlations. Although both the spatial and temporal perspectives have been considered in modeling, most existing works have ignored complex temporal correlations or underlying spatial similarity. In this paper, we identify the unique spatiotemporal correlation of urban metro flow, and propose an attention-based deep spatiotemporal network with multi-task learning (ADST-Net) at a citywide level to predict the future flow from historical observations. ADST-Net uses three independent channels with the same structure to model the recent, daily-periodic and weekly-periodic complicated spatiotemporal correlations, respectively. Specifically, each channel uses the framework of residual networks, the rectified block and the multi-scale convolutions to mine spatiotemporal correlations. The residual networks can effectively overcome the gradient vanishing problem. The rectified block adopts an attentional mechanism to automatically reweigh measurements at different time intervals, and the multi-scale convolutions are used to extract explicit spatial relationships. ADST-Net also introduces an external embedding mechanism to extract the influence of external factors on flow prediction, such as weather conditions. Furthermore, we enforce multi-task learning to utilize transition passenger flow volume prediction as an auxiliary task during the training process for generalization. Through this model, we can not only capture the steady trend, but also the sudden changes of passenger flow. Extensive experimental results on two real-world traffic flow datasets demonstrate the obvious improvement and superior performance of our proposed algorithm compared with state-of-the-art baselines. [ABSTRACT FROM AUTHOR]