Person re-identification is essential to intelligent video analytics, whose results affect downstream tasks such as behavior and event analysis. However, most existing models only consider the accuracy, rather than the computational complexity, which is also an aspect to consider in practical deployment. We note that self-attention is a powerful technique for representation learning. It can work with convolution to learn more discriminative feature representations for re-identification. We propose an improved multi-scale feature learning structure, DM-OSNet, with better performance than the original OSNet. Our DM-OSNet replaces the 9×9 convolutional stream in OSNet with multi-head self-attention. To maintain model efficiency, we use double-layer multi-head self-attention to reduce the computational complexity of the original multi-head self-attention. The computational complexity is reduced from the original O((H×W)2) to O(H×W×G2). To further improve the model performance, we use SpCL to perform unsupervised pre-training on the large-scale unlabeled pedestrian dataset LUPerson. Finally, our DM-OSNet achieves an mAP of 87.36%, 78.26%, 72.96%, and 57.13% on the Market1501, DukeMTMC-reID, CUHK03, and MSMT17 datasets.

With the rapid popularization of intelligent terminals and the explosive growth of wireless communication service demand, future mobile communication technology will face many challenges. Non-orthogonal multiple access (NOMA) technology for 5G can provide many connections and effectively improve the frequency spectrum and energy efficiency compared to traditional orthogonal multiple access technologies. Therefore, in recent years, NOMA technology has become one of the research hotspots of numerous scholars. However, the resource allocation problem in the NOMA system, as a high-dimensional nonlinear programming problem, has not been well studied. In addition, the particle swarm optimization algorithm can also effectively find the optimal solution for complex and constrained problems. Still, at the same time, it is easy to fall into local optimal defects. In this context, we decouple the high-dimensional nonlinear programming problem to maximize system energy efficiency into sub-problems: subchannel and power allocation. Firstly, a low-complexity greedy algorithm based on the principle of worst-case subchannel priority matching is proposed to solve the subchannel assignment problem. In addition, we further apply the modified particle swarm optimization algorithm to allocate power to the NOMA downlink system, aiming to improve the energy efficiency of the communication system as much as possible under the premise of ensuring the quality of service (QoS). Simulation results show that our proposed scheme has low complexity and can significantly improve the energy efficiency of the NOMA system and achieve better user fairness.

No abstract