Implement Real Time Dense Optical Flow Inference on Jetson Nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

With cuda support OpenCV library it cost 3 seconds to infer a flow from two frames, my goal is to build a python program that can calculate dense optical about 5 ~ 10 FPS.

Any suggestion about API or SDK I can start with?

3

Hi,

Unfortunately, we don’t have a dense optical flow sample for Nano.

Have you maximized the device performance when running OpenCV?

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

More, maybe you can downscale the image to see if any improvemnt if this is acceptable.

Thanks.

5

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.