Hello, we would like to build a low-power computer vision application.
It seems like a good way of doing this is by using HVX to stream the camera to the L2 (as in the "camera streaming" example), then using the Hexagon DSP to run a computer vision algorithm on the image data.
Are there any examples of HVX camera streaming + object detection? What is the recommended way to achieve this?
I see that the Snapdragon Neural Processing Engine (NPE) supports high-level APIs for running neural networks on the Hexagon DSP. Is it possible to use the HVX camera stream as an input the NPE without hitting the CPU? Thank you for your help.