Obj_user_meta_list shows None on nvinferserver source pad_probe

[abdo.babukr1]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU)
Jetson Orin AGX Dev Kit
• DeepStream Version
6.1.1
• JetPack Version (valid for Jetson only)
5.0.2

We are using nvinferserver for a primary detector and a secondary inference.

On the secondary config file we include

postprocess {
    other {}
  }

and

input_control {
  process_mode: PROCESS_MODE_CLIP_OBJECTS
  operate_on_gie_id: 5
  operate_on_class_ids: [1,2]
  async_mode: true
  object_control {
    bbox_filter {
      min_width: 32 
      min_height: 32
    }
  }
}
output_control {
  output_tensor_meta: true
}

We are probing the source pad of the secondary inference and are able to get the class id and the width/height.

obj_user_meta_list is showing None and we expect to get the [1x512] raw tensor. What could be wrong? We are getting class 1 for example, and it should be operating on class 1 but we aren’t getting the output tensor from frame_meta_list> obj_meta_list> obj_user_meta_list

[fanzh]

please refer to nvinferserver pgie+sgie sample deepstream-infer-tensor-meta-test.

[abdo.babukr1]

Thank you @fanzh the example was in fact exactly what we were trying to accomplish, and the solution was to use a large batch-size.

Would it be possible to achieve the same thing with performing the post processor on the source pad of the primary inference plugin instead of using a post processor library? I can open another ticket if needed.

Thanks,
Abdo

[fanzh]

if meet new issue, please open a new topic, Thanks!

[abdo.babukr1]

Yes this is a new issue, and instead of a post processor lib we are using the source pad to do the post processing like the example from the python deep stream repo.

Postprocessor performed on Primary Inference Source Probe not allowing Secondary Inference to Process results

• Hardware Platform (Jetson / GPU) Jetson Orin AGX Dev Kit • DeepStream Version 6.1.1 • JetPack Version (valid for Jetson only) 5.0.2 We have a post processing library that works with a secondary inference, we are able to access the output tensor of the secondary inference in a source probe. We are using a python application. The issue is the obj_meta.obj_user_meta_list shows as None when implementing the exact post processor on the primary inference source probe instead of using a postpr…

Thanks,
Abdo