I am using the Jetson AGX Xavier board with Ubuntu 18.04.
Also, I am using YoloV3 with an ethernet camera and ROS to classify obstacles (darknet_ROS).
I have a question about the number of FPS. In fact, I change the reolution of the input image, 1920x1080 and 640x480, and I get the same fps, about 10 FPS for both of them. I want to konw why I get the same FPS for the tow resolutions? I expected more FPS for the 640x480 resolution.
I think (but I could be wrong) Yolo_v3 resamples all input pictures to the size it’s trained on, so detection speed should be the same (except for the small time taken to resample.)
It’s been a while since I used it, but at least it’s something to look into!
What’s kind of camera, USB or CSI?
It is an ethernet camera (axis F1015).
About the resampling of input pictures, I changed the input layer size to get more FPS, so I got it but the fps still same for both of resolutions. The following table presents the results:
Also, I got almost 100% occupation for GPU for all tests and for different resolutions:
You may need to verify fps without Yolo to narrow the root cause.
I already did that. I checked the fps from the output of camera and it worked fine. Also, I have a different FPS for yolov2-tiny:
Note that the occupation of CPU is different for the two resolution, for the VGA resolution it is as below:
And for the Full HD:
Did you boost the system clocks?
sudo nvpmodel -m 0 sudo jetson_clocks
Yes, I did it.
In fact, now, I try to find an explanation about the same FPS for both resolutions when I used YoloV3 (with darknet_ros framework ).
Then I think it could be YoloV3’s problem.
This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.