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]

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]

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]

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]

Deep learning has become an essential technique in image steganography. Most of the current deep-learning-based steganographic methods process digital images in the spatial domain. There are problems such as limited embedding capacity and unsatisfactory visual quality. To improve capacity-distortion performance, we develop a steganographic method from the frequency-domain perspective. We propose a module called the adaptive frequency-domain channel attention network (AFcaNet), which makes full use of the frequency features in each channel by a fine-grained manner of assigning weights. We apply this module to the state-of-the-art SteganoGAN, forming an Adaptive Frequency High-capacity Steganography Generative Adversarial Network (AFHS-GAN). The proposed neural network enhances the ability of high-dimensional feature extraction through overlaying densely connected convolutional blocks. In addition to this, a low-frequency loss function is introduced as an evaluation metric to guide the training of the network and thus reduces the modification of low-frequency regions of the image. Experimental results on the Div2K dataset show that our method has a better generalization capability compared to the SteganoGAN, with substantial improvement in both embedding capacity and stego-image quality. Furthermore, the embedding distribution of our method in the DCT domain is more similar to that of the traditional method, which is consistent with the prior knowledge of image steganography. [ABSTRACT FROM AUTHOR]