How do I provide AGX speed to run YOLO-v4

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Xavier
1

HI！
AGX executes YOLO-v4 more slowly than NVIDIA T4.
AGX takes 140ms to execute one frame, and T4 takes 40ms to execute.
Then I executed “sudo NVPmodel-m 0” at a speed of 80ms.
Is there any other way to make AGX faster? I’ve heard friends say that AGX is stronger than T4.

3

Hi,

Please try the following command to maximize the Xavier performance:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

Thanks.

4

Hi, here’s what I did,

5

Thanks for the feedback.

We are checking this internally and will share more information with you later.

6

Hi,

Please try our sample listed below:

We can get around 57.74 fps on Xavier with Deepstream pipeline.
This is end-to-end performance including camera, inference and recording.

Thanks.