OpenCV is very Slow

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Heloo
I’am working on a deep learning project to preform emotion recognition and sleep detection using the jetson tx2 board. i’am using openCV to open the camera and capture a frame from the live stream the forward this the opencv function cv2.detectmultiscale to preform face detection using the haarcascade face detection model but there is a huge delay (about 6 seconds)

i can’t find any solution to this problem as any other model like mtcnn require openCv-pyhton to be installed which is not supported by the board as i found in some forms so my question is, is there any way to do this task in an efficient way such that i can prevent this delay ?

2

Hi,
Please try sudo jetson_clocks
OpenCV is CPU-based implementation and you should see improvement to run CPU at max clocks.

We have an optimal solution of running deep learning inference on Jetson platforms:
https://docs.nvidia.com/metropolis/index.html

Please install DeepStream SDK through SDKManager and give it a try.

3

thanks for replying

i tried the first solution but it didn’t make the performance better

regarding the second solution i need to know if DeepStream is compatible with python so that i can open the jetson tx2 camera and capture a frame and feed this frame to CNN ?

any tutorial or instruction to preform such a task would be appreciated .

Thanks in advance

4

Hi,
We have python samples in

Please give it a try.