排序方式: 共有3条查询结果,搜索用时 0 毫秒
1
1.
2.
黄宏宇 《高技术通讯(英文版)》2009,15(4):363-368
This paper studies a routing protocol used in the application of collecting real-time traffic information using mobile vehicles to monitor traffic status. The biggest challenge of vehicular ad hoc network routing protocol lies in the mismatch between the direction of moving vehicles and routed data. In order to deal with this mismatch, a mesh based routing protocol with a two-tier network architecture is proposed. By using mesh nodes deployed at intersections, data can be routed through an optimized path which can improve the delivery ratio and reduce the consumption of network resources. The simulation uses the mobility model processed from the GPS data of taxis running in Shanghai urban areas. The result shows that the proposed protocol outperforms the existing flooding protocol. 相似文献
3.
图像自动生成一直以来都是计算机视觉领域的一项重要挑战,其中的文本到图像的生成更是图像生成领域的重要分支。随着深度学习技术迅猛发展,生成对抗网络的出现使得图像生成领域焕发生机,借助生成对抗网络能够生成较为生动且多样的图像。本文将自注意力机制引入生成对抗网络,提出GAN-SelfAtt以提升生成图像的质量。同时,使用WGAN、WGAN-GP 2种生成对抗网络框架对GAN-SelfAtt进行实现。实验结果表明,自注意力机制的引入能够提高生成图像的清晰度,这归功于自注意力机制弥补了卷积运算中只能计算局部像素区域内的相关性的缺陷。除此之外,GAN-SelfAtt在训练时有着更好的稳定性,避免了原始生成对抗网络中的模式坍塌问题。 相似文献
1