Maintain custom parameters for a tracked object across inference iterations

[artemchepurnoy]

**• Hardware Platform (Jetson / GPU): Jetson, NVIDIA Xavier NX 8GB
**• DeepStream Version: DeepStream 6.2
• JetPack Version (valid for Jetson only): JetPack 5.1.1
• TensorRT Version: TensorRT 8.5.2
• Issue Type( questions, new requirements, bugs): question

My program determines some custom parameters for tracked objects.
I need to save them and maintain across inference iterations.

I tried to use obj_user_meta_list to save the parameters. For this I:

• acquire an instance of NvDsUserMeta by calling user_meta = nvds_acquire_user_meta_from_pool(batch_meta)
• initialize user meta
• add the user meta to an object meta by calling nvds_add_user_meta_to_obj(obj_meta, user_meta)

It works OK but only until I get a new frame.
After getting a new frame obj_meta->obj_user_meta_list == NULL always returns true, so I cannot access the user_meta.

What should I do to access the previously saved user_meta after getting a new frame?

[yuweiw]

The meta is bound with the current frame. So if you want to get the previously saved user_meta, you can bind this to the new frame by yourself and get that.

[artemchepurnoy]

Ok. And what is the best practice to do this (I mean bind user_meta from objects of the previous frame to objects of the next frame)?
Should I

1. get metadata of the previous frame using nvds_get_nth_frame_meta()
2. iterate through all objects at the previous frame
3. find the same objects at the new frame
4. copy user_meta from objects at the previous frame to the same objects at the new frame
   ???

It sounds ineffective. Is there a more optimal way?

[yuweiw]

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

You can create a list of the meta of all frames by yourself and copy the metas in it. When you need that, you can search and obtain it in the list.