How can I track and detect objects in real time with YOLOv5 and Jetson Nano?


As the title says, I am trying to track objects using YOLOv5 with a custom dataset on my Jetson Nano. I have YOLOv5 working for the custom dataset, and I tried to use DeepSORT for this. The problem is that DeepSORT takes a long time, and I am getting 4-5 FPS for the whole process.
Is there any other method I can use on my Jetson Nano for this? I need to track with the camera and video, and 4-5 FPS is just not enough :(



Have you maximized the device performance first?

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

If the performance doesn’t meet the requirement after boosting, please check the GPU resources with tegrastats and share it with us.

$ sudo tegrastats


1 Like

I’m still getting 4-6 FPS, 45.2ms for YOLO inference, and 99.6 for DeepSORT.
There is my tegrastats results:


Are you running the tegrastats and the YOLO inference at the same time?
Based on the log, the GPU utilization is 0% (GR3D_FREQ).


My bad, now i’m running the tegrastats and the YOLO inference at the same time. There is my results:

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