People tracking help needed

• Hardware Platform (Jetson / GPU)
Jetson Xavier NX and AGX
• DeepStream Version
• JetPack Version (valid for Jetson only)
• Issue Type( questions, new requirements, bugs)

Hello nvidia family and friends,

I’m working on a project where I need to control a camera mounted on a Gimbal to follow people when a person is detected.
It works well when there is only one person on the screen, the camera follows that person and keeps it in the center of the image.
My problem is when there is two or more people. Even if these two persons remain at the same place, the camera will move towards one, then switch to the other, then back to the first, etc, not being able to stay on one of them.

I tried to only detect one person by adding a is_first_detected_object boolean but it fails.
I tried using the nvtracker to use an ID assigned to a detected object, but I guess I’m not good at using it correctly as it fails too.

My goal is to follow one person for a certain amount of time, then after some time, if there is a second person in the field of view, I would like to switch the focus to that next person.

If someone could help me here, give me an idea I didn’t think about or any kind of hint, I would greatly appreciate.
I’m coding in python and I’m using a yolo to only detect people.


What does this mean?

With nvtracker enabled, object_id in NvDsObjectMeta struct can reflect the identified object.

It means that I was able to get the ID assigned to the detected object, but I couldn’t find a smart way to use it to help me.
I tried putting the IDs in a set and only track the person with the smallest ID in the set, but unsuccessfully apparently.
My problem is on the algorithmic side :-p

There is no update from you for a period, assuming this is not an issue any more.
Hence we are closing this topic. If need further support, please open a new one.

Which algorithm? Our tracking algorithm?

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