To address the problems of missing detection and low detection accuracy of the existing helmet wearing detection algorithms for small and crowded targets detection, this paper proposes a helmet wearing detection method based on improved YOLOv5 and transfer learning. First, different from the default priori frame that is not suitable for the task, we use the K-means algorithm to cluster the suitable priori frame size for the detection task. Then, in the back of the feature extraction network, we introduce a spatial channel mixed attention module to strengthen the learning of relevant weights and suppress the weights of irrelevant backgrounds, respectively. Further, we improve the judgment metric of the non-maximum-suppression (NMS) algorithm in the post-processing stage of YOLOv5 to reduce the phenomenon of false deletion and missing of prediction boxes. After that, the proposed network is trained based on the strategy of transfer learning, which can overcome the scarcity of limited existing data sets and improve the generalization ability of the model. Finally, we build a cascade judgment framework for helmet wearing deployed in visual sensor networks. The experimental results show that our proposed method improves the average accuracy (IOU=0.5) to 93.6%, which is 5% higher than the original model in the helmet wearing data set. The proposed model also outperforms other state-of-the-art algorithms by obviously improving the accuracy of helmet wearing detection in the construction scenarios.

Low-dose single-photon emission computed tomography (SPECT) imaging can reduce the radiation damage to human bodies caused by radioactive tracers, and hence it is becoming more and more important in clinical practice. In a SPECT system, low-dose imaging can be achieved by acquiring projection data of sparse-view. The sparse-view projection data, if directly reconstructed by conventional iterative reconstruction methods, will inevitably lead to severe ray artifacts in the image domain. Existing clinical reconstruction methods usually introduce specific regularization to the optimization model to suppress ray artifacts. However, this type of methods may not adapt to projection data with various dosage, and the form of regularization heavily depends on prior knowledge. A novel neural network architecture is proposed to learn the mapping from the sparse-view projection data to the full-view projection data. The projection data of missing view angle is synthesized by the proposed neural network to improve the quality of reconstructed images. Numerical experiments show that, compared with the traditional iterative reconstruction method, the SSIM of the reconstructed image is increased by 59%, the NMSE is reduced by 67%, and the PSNR is increased by 2.48 dB. Therefore, the proposed method can better improve the image quality of sparse-view projection data.

To address the problems of camouflage and real-time detection of the traditional Android malware detection methods, a new Android malware detection method based on heterogeneous information networks is proposed. By modeling the Android entities and relationships nodes and edges, respectively, in a heterogeneous information network, two network representation learning models are designed, including the meta-structure attention network representation learning and the incremental learning models. First, the meta-structure attention network representation learning model is used for intra-sample node embedding, and the embedded nodes and labels are input to a deep neural network for training. Then, the incremental learning model is used for learning the extra-sample node embeddings. The top-k algorithm is used to aggregate neighboring nodes within the heterogeneous information network, and the embedded node to be detected is input to the trained deep neural network for detection. Experimental results show that the F₁ value of the proposed method is 97.5%, the accuracy rate is 96.7%, and the average detection time is 3.7 ms, which are better than the existing methods, demonstrating the effectiveness of the proposed method for dealing with Android malware camouflage and for real-time Android malware detection.

Recent StyleGAN-based face swapping methods have been able to generate very realistic high-resolution face swapping results, but they are often plagued by the challenge of maintaining various attributes (such as expression, pose, and illumination). One reason is that these methods usually focus on the latent codes of facial semantic features corresponding to the $\mathcal{W}/\mathcal{W}+$ space, and latent codes in these spaces are often highly entangled. To address this issue, we propose a new method, LDSwap. for disentangling and re-fusing latent codes in the StyleSpace. The semantic-related latent code disentangling module (SLDM) we propose can successfully achieve facial semantic feature exchange and reorganization by disentangling latent codes. In addition, we propose a channel-split adaptive feature fusion module (CAFF) that adaptively learns and refuses spatial information in the target image. This module can learn spatial features from the target image without interference from the features of the target face region. Through qualitative and quantitative evaluation, we demonstrate that LDSwap shows significant improvements over three state-of-the-art methods in maintaining the appearance of semantic features.