CUDA核心用于实时处理。 [英] CUDA Cores for real time processing.
问题描述
我正在考虑一个需要使用CUDA核心来对Kinect数据进行实时分析的应用程序。在理想的情况下，我的应用程序将位于USB接口和Kinect SDK之间，这样繁重的计算就可以运行
并输入Kinect SDK，就像它来自USB一样。这是可能的，还是我将无法创建虚拟USB接口，捕获Kinect数据流，使用CUDA核心处理它们，然后将它们作为虚拟USB输出提供给SDK。
如果需要虚拟化，我在哪里可以找到所需的所有技术规范来欺骗SDK和Kinect以确信它们是直接连接的？
实现它。
I'm considering an application that would require the use of CUDA cores to do real time analysis of the Kinect data. In the ideal scenario, my application would site between the USB interface and the Kinect SDK so that the heavy computations could be run and fed to the Kinect SDK as if it were coming from the USB. Is this possible, or would I be stuck creating a virtual USB interface, capture the Kinect data streams, process them with CUDA cores and then provide them to the SDK as a virtual USB output. If it would need to be virtualized, where can I find all of the technical specifications required to spoof the SDK and the Kinect into believing they were directly connected?
Make it happen.
推荐答案
没有办法输入Kinect管道。我们直接从传感器提取数据以生成深度/体/等数据。您CUDA处理将使用SDK数据并对其执行某些操作。
There is no way to feed into the Kinect pipeline. We pull data directly from the sensor to generate depth/body/etc data. You CUDA processing would consume the SDK data and do something with that.
这篇关于CUDA核心用于实时处理。的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
