Using Jetson Inference for object counting with live video from UAV

Hi, I’m working on a project to analyse UAV videos and counting the total objects identified for the video (so accumulative count, not just count per frame) in real-time using Jetson Nano (which I’m thinking can upgrade later).
I’ve looked at the inference tutorial which only only detects objects but doesn’t track or use other algorithms for counting. Any existing methods that already implements counting that anyone can kindly point me to?

I’ve had a look at possibility of Deepstream but it seems a steep learning curve. I’ve come across this but it seems to be on the 5.1 version but currently is 6.1 so not sure where to go Can I get the accumulate number of objects passed ROI?.

Any help appreciated! Thanks


Tracker is essential for object counting.
If can help you to identify the same object cross different frames.
(Or you may add the same object for multiple times)

Unfortunately, we don’t have an available counting example.
But you can start with Deepstream or Jetson-infernece.


Hello, thanks for the reply. Does the Jetson inference include tracking?


No. But you can update the source with the tracker supported by OpenCV.


Thank you. I’ll have a look at the OpenCV supported tracker, will come back if have any questions.

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