Performance optimize - openCV, numpy, scipy

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Dear,

I’d like to know how to optimize the performance at Jetson TX2.
I mainly use OpenCV, Numpy, Scipy functions for image processing.

Please let me know some tips for optimization.

  1. how to optimize OpenCV at Jetson environment

    • by compile? by using or not using specific function?

  2. how to optimize array operations using Numpy and Scipy.

    • by using some Jetson specific command or compile option for Numpy package?

Thanks in advance

CHJung

2

chjung90,

There is no correct answer for you question since we don’t know what sample you are using.

From the point of system side, you could use nvpmodel and jetson_clock.sh to enhance the performance since these tools will use higher cpu/gpu/emc frequency.

As for OpenCV, please note the the default opencv in jetpack does not have cuda or gpu acceleration. So if you want to use that, please try to build openCV lib from scratch.